Energy system control method and device considering wind-solar power generation power, computer equipment, readable storage medium and program product

Abstract

The invention relates to an energy system control method and device considering wind and light power generation power, computer equipment, a computer readable storage medium and a computer program product, relates to the field of system control, and can improve the control real-time performance of an energy system. The method comprises the following steps: acquiring wind and light power generation data; the wind-solar power generation data comprises current wind-solar power generation power data and weather environment data; extracting the time sequence characteristics of the wind-solar power generation data through the prediction model, and the correlation between the current wind-solar power generation power data and the weather environment data, and obtaining the predicted wind-solar power generation power according to the time sequence characteristics and the correlation; acquiring operation states of a plurality of energy units of the energy system; controlling a plurality of energy units under a control instruction in a control instruction set based on the predicted wind-solar power generation power and the operation state, and determining operation resources consumed by an energy system; and determining a target control instruction for the plurality of energy units from the control instruction set by taking the operation resource minimization as a target.

Description

Technical Field

[0001] The present application relates to the technical field of system control, and particularly to an energy system control method, device, computer device, computer-readable storage medium, and computer program product considering wind and solar power generation. Background Art

[0002] In the context of large-scale access of renewable energy, the electricity-hydrogen coupled energy system faces significant challenges. The output of wind and photovoltaic power generation has strong volatility and uncertainty, bringing pressure to the stable operation and economic dispatching of the system. Although existing research has attempted to improve the power prediction accuracy by using traditional machine learning, deep learning, and non-parametric methods, or enhance the flexibility and responsiveness of the system through model predictive control (MPC) and parameter adaptive sliding mode predictive control (PASMPC), it is still difficult to achieve an effective balance between prediction accuracy and real-time response ability, which has become a key technical problem restricting the efficient operation of the system. Summary of the Invention

[0003] Based on this, it is necessary to provide an energy system control method, device, computer device, computer-readable storage medium, and computer program product considering wind and solar power generation for the above technical problems.

[0004] In a first aspect, the present application provides an energy system control method considering wind and solar power generation, including:

[0005] Obtain wind and solar power generation data; the wind and solar power generation data includes current wind and solar power generation power data and weather environment data;

[0006] Extract the time series features of the wind and solar power generation data and the correlation between the current wind and solar power generation power data and the weather environment data through a prediction model, and predict the wind and solar power generation power according to the time series features and the correlation to obtain the predicted wind and solar power generation power;

[0007] Obtain the operating states of multiple energy units of the energy system;

[0008] Based on the predicted wind and solar power generation power and the operating states, control the multiple energy units under the control instructions in the control instruction set, and determine the operating resources consumed by the energy system;

[0009] With the goal of minimizing the operating resources, determine the target control instructions for the multiple energy units from the control instruction set.

[0010] In one embodiment, based on the predicted wind and solar power generation and the operating state, controlling the multiple energy units under the control instructions in the control instruction set, and determining the operating resources consumed by the energy system includes:

[0011] Determine a first operating state and a second operating state in the operating state; the first operating state is the operating state that occurs at the next moment of the second operating state;

[0012] According to the first operating state and the second operating state, determine the working trend of the energy system;

[0013] Based on the predicted wind and solar power generation and the working trend, control the multiple energy units under the control instructions in the control instruction set, and determine the operating resources consumed by the energy system.

[0014] In one embodiment, the operating resources include the operating resources consumed by each of the energy units; taking minimizing the operating resources as the goal, determining the target control instructions for the multiple energy units from the control instruction set includes:

[0015] Determine the energy units according to the operating resources consumed by each of the energy units;

[0016] Obtain the operating parameters of the energy unit under the control instruction, and compare the operating parameters with a preset parameter threshold to obtain a comparison result; the comparison result is used to indicate the control priority of the energy unit;

[0017] Based on the control priority, taking minimizing the operating resources as the goal, determine the target control instructions for the multiple energy units from the control instruction set; the target control instructions include turning on and off the energy unit.

[0018] In one embodiment, the prediction model includes an encoder network and a decoder network;

[0019] Extracting the time series features of the wind and solar power generation data through the prediction model, and the correlation between the current wind and solar power generation power data and the weather environment data, and predicting the wind and solar power generation power according to the time series features and the correlation to obtain the predicted wind and solar power generation power, includes:

[0020] Input the wind and solar power generation data into the encoder network to extract the encoding result representing the time series features;

[0021] Acquire weather forecast data from the weather environment data for a preset time period, and input the encoding result and the encoding feature formed by the weather forecast data into the decoder network, determine the intermediate feature corresponding to the encoding feature by the intermediate layer of the decoder network, and obtain a fused feature by fusing the encoding feature and the intermediate feature; the fused feature indicates the correlation between the current wind and solar power generation data and the weather environment data;

[0022] The fused features are input into the fully connected layer of the decoder network, and the predicted wind and solar power generation is output.

[0023] In one embodiment, the prediction model is obtained by the following steps:

[0024] Acquire historical data of wind and solar power generation, and classify the historical data of wind and solar power generation according to preset classification conditions to obtain historical data of wind and solar power generation in multiple categories;

[0025] For each of the categories of wind and solar power generation historical data, model training is performed according to the category of wind and solar power generation historical data to obtain a prediction sub-model corresponding to the category of wind and solar power generation historical data;

[0026] Integrate a plurality of the prediction sub-models to obtain the prediction model;

[0027] After obtaining the wind and solar power generation data, the method further includes:

[0028] Wind and solar power generation data belonging to different categories are determined, and the wind and solar power generation power corresponding to each category of wind and solar power generation data is predicted according to the prediction sub-models corresponding to the wind and solar power generation data of different categories to obtain the predicted wind and solar power generation power.

[0029] In one embodiment, the model training is performed according to the historical data of wind and solar power generation of the category to obtain the prediction sub-model corresponding to the historical data of wind and solar power generation of the category, including:

[0030] Inputting the wind-solar power generation historical data into the prediction sub-model to be trained, determining the hidden state information corresponding to the time series features in the wind-solar power generation historical data by the prediction sub-model to be trained, and removing the weighted processing result of the hidden state information from the wind-solar power generation historical data to obtain intermediate training data;

[0031] Determine the degree of change of the weather environment corresponding to the weather environment data of the wind and solar power generation historical data, and use the intermediate training data and the degree of change of the weather environment as target training data;

[0032] Determining, by the prediction sub-model to be trained, the training prediction of wind and solar power generation power according to the target training data;

[0033] Adjust the parameters of the prediction sub-model according to the difference between the predicted wind and solar power generation and the actual wind and solar power generation in the historical wind and solar power generation data, and update the weights corresponding to the hidden state information and the information to be hidden;

[0034] Return to execute the step of removing the weighted processing result of the hidden state information from the historical wind and solar power generation data to obtain intermediate training data until the preset training conditions are met, and obtain the prediction sub-model corresponding to the category.

[0035] In a second aspect, the present application also provides an energy system control device considering wind and solar power generation, including:

[0036] A wind and solar power generation data acquisition module for acquiring wind and solar power generation data; the wind and solar power generation data includes current wind and solar power generation power data and weather environment data;

[0037] A wind and solar power generation power prediction module for extracting the time series characteristics of the wind and solar power generation data and the correlation between the current wind and solar power generation power data and the weather environment data through a prediction model, and predicting the wind and solar power generation power according to the time series characteristics and the correlation to obtain the predicted wind and solar power generation power;

[0038] An energy unit operating state acquisition module for acquiring the operating states of multiple energy units of the energy system;

[0039] An operating resource determination module for controlling the multiple energy units under the control instructions in the control instruction set based on the predicted wind and solar power generation power and the operating states, and determining the operating resources consumed by the energy system;

[0040] A target control instruction determination module for determining the target control instructions for the multiple energy units from the control instruction set with the goal of minimizing the operating resources.

[0041] In a third aspect, the present application also provides a computer device, including a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0042] Acquire wind and solar power generation data; the wind and solar power generation data includes current wind and solar power generation power data and weather environment data;

[0043] Extract the time series characteristics of the wind and solar power generation data and the correlation between the current wind and solar power generation power data and the weather environment data through a prediction model, and predict the wind and solar power generation power according to the time series characteristics and the correlation to obtain the predicted wind and solar power generation power;

[0044] Obtain the operating states of multiple energy units of the energy system;

[0045] Based on the predicted wind and solar power generation and the operating states, control the multiple energy units under the control instructions in the control instruction set, and determine the operating resources consumed by the energy system;

[0046] With the goal of minimizing the operating resources, determine the target control instructions for the multiple energy units from the control instruction set.

[0047] In a fourth aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0048] Obtain wind and solar power generation data; the wind and solar power generation data includes current wind and solar power generation power data and weather environment data;

[0049] Extract the time series features of the wind and solar power generation data and the correlation between the current wind and solar power generation power data and the weather environment data through a prediction model. According to the time series features and the correlation, predict the wind and solar power generation power to obtain the predicted wind and solar power generation power;

[0050] Obtain the operating states of multiple energy units of the energy system;

[0051] Based on the predicted wind and solar power generation and the operating states, control the multiple energy units under the control instructions in the control instruction set, and determine the operating resources consumed by the energy system;

[0052] With the goal of minimizing the operating resources, determine the target control instructions for the multiple energy units from the control instruction set.

[0053] In a fifth aspect, the present application also provides a computer program product, including a computer program. When the computer program is executed by a processor, the following steps are implemented:

[0054] Obtain wind and solar power generation data; the wind and solar power generation data includes current wind and solar power generation power data and weather environment data;

[0055] Extract the time series features of the wind and solar power generation data and the correlation between the current wind and solar power generation power data and the weather environment data through a prediction model. According to the time series features and the correlation, predict the wind and solar power generation power to obtain the predicted wind and solar power generation power;

[0056] Obtain the operating states of multiple energy units of the energy system;

[0057] Based on the predicted wind and solar power generation and the operating state, control the multiple energy units under the control instructions in the control instruction set, and determine the operating resources consumed by the energy system;

[0058] With the goal of minimizing the operating resources, determine the target control instructions for the multiple energy units from the control instruction set.

[0059] The above energy system control method, device, computer device, computer-readable storage medium, and computer program product that consider wind and solar power generation obtain wind and solar power generation data; the wind and solar power generation data includes current wind and solar power generation power data and weather environment data; extract the time series characteristics of the wind and solar power generation data and the correlation between the current wind and solar power generation power data and the weather environment data through a prediction model, and predict the wind and solar power generation power according to the time series characteristics and the correlation to obtain the predicted wind and solar power generation power; obtain the operating states of multiple energy units of the energy system; based on the predicted wind and solar power generation and the operating state, control the multiple energy units under the control instructions in the control instruction set, and determine the operating resources consumed by the energy system; with the goal of minimizing the operating resources, determine the target control instructions for the multiple energy units from the control instruction set. In this application, by obtaining the real-time power data of wind and solar power generation and the environment-related data, and using the prediction model to deeply mine the time series characteristics and internal associations of these data, high-precision prediction of future wind and solar power generation power is realized; on this basis, combined with the real-time operating states of multiple energy units of the energy system, dynamic optimization decisions are made with the goal of resource minimization, reducing the curtailment rate of wind and solar power and the standby energy usage cost; at the same time, the dynamic scheduling strategy reduces the frequent start-stop and overcharge-discharge of energy equipment, extends the service life of the equipment, and improves the real-time performance of the control of the energy system while ensuring the reliability of the system. Description of the Drawings

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments of the present application or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0061] Figure 1 It is a framework diagram of an energy system in an embodiment;

[0062] Figure 2 It is a flowchart of an energy system control method that considers wind and solar power generation in an embodiment;

[0063] Figure 3 It is a flowchart of reinforcement learning in an embodiment;

[0064] Figure 4 It is a schematic diagram of the training process of a prediction model in an embodiment;

[0065] Figure 5 It is a schematic diagram of the process of an energy system control method considering wind and solar power generation in another embodiment;

[0066] Figure 6 It is a structural block diagram of an energy system control device considering wind and solar power generation in an embodiment;

[0067] Figure 7 It is an internal structure diagram of a computer device in an embodiment. Specific implementation manners

[0068] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0069] The energy system control method considering wind and solar power generation provided by the embodiments of the present application. In one embodiment, the energy control system can be as Figure 1 shown, and it consists of energy units such as wind turbines, photovoltaics, an energy conversion unit (including a CHP <Combined Heat and Power> unit and a gas boiler), P2G (Power-to-Gas) equipment, a storage battery, and a hydrogen storage tank to form an ETHC-IES (Integrated Energy System) energy system, which is interconnected through a natural gas network. The CHP unit converts natural gas into electric and heat energy, and the gas boiler converts natural gas into heat energy. The energy system can store the surplus electric energy of renewable energy through a storage battery, or produce hydrogen through an electrolyzer and store it in the hydrogen storage tank. After meeting the hydrogen load, the hydrogen storage tank can convert more than 50% of its capacity of hydrogen into electric and heat energy through a hydrogen fuel cell to supplement the load, and the storage battery supplies power to supplement when the power is insufficient.

[0070] In one embodiment, as Figure 2 shown, a method for controlling an energy system considering wind and solar power generation is provided. In this embodiment, the method is exemplified by being applied to a terminal. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is realized through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0071] Step S202, obtaining wind and solar power generation data; the wind and solar power generation data includes current wind and solar power generation power data and weather environment data.

[0072] Among them, the wind-solar power generation data is a composite time-series data set containing the output power on the equipment side and the meteorological variables on the environmental side. By continuously acquiring and updating such data, it can provide basic support for subsequent power prediction, operation status evaluation, and energy scheduling. The current wind-solar power generation power data can be the real-time active power values output by wind turbines and photovoltaic modules at a certain moment, which is used to reflect the power generation capacity of renewable energy under the current meteorological conditions and energy system configuration. The weather environment data is the external meteorological condition data that directly affects the output of wind-solar power generation, including but not limited to one or more of wind speed, wind direction, solar irradiance, temperature, humidity, air pressure, cloud cover, etc.

[0073] Exemplarily, the terminal performs the acquisition task of wind-solar power generation data by communicating with the equipment on the station side of the energy system. The terminal establishes a communication connection with the wind turbine controller and the photovoltaic inverter, obtains the real-time power values of wind power output and photovoltaic power output at the current moment, and marks them with timestamps to form standardized current wind-solar power generation power data. At the same time, environmental perception sensors are deployed in the energy system, and the terminal forms weather environment data by acquiring, regularizing, and recording the data of the environmental perception sensors.

[0074] Optionally, the terminal can perform outlier processing, normalization, data dimensionality reduction, or other forms of preprocessing on the acquired wind-solar power generation data to improve the stability and accuracy of subsequent model recognition.

[0075] Step S204, extract the time-series features of the wind-solar power generation data and the correlation between the current wind-solar power generation power data and the weather environment data. According to the time-series features and the correlation, predict the wind-solar power generation power to obtain the predicted wind-solar power generation power.

[0076] Among them, the time-series features can be the statistical trends, periodicity, volatility, hysteresis, etc. formed by the wind-solar power generation data over time, which can characterize the evolution pattern of the output of wind energy and solar energy renewable energy in the time dimension, such as sudden changes in meteorological changes. The correlation can be the correlation between the wind-solar power generation power and the weather environment data (such as wind speed, irradiance, etc.), which is used to reveal how weather changes affect the power generation capacity. The prediction model can be a machine learning model with time-series modeling and feature fusion capabilities, which can be modeled based on a deep learning framework, such as the LSTM (Long Short-Term Memory) or Transformer architecture. Its main responsibility is to numerically predict the wind-solar power generation power in the future period.

[0077] Exemplarily, based on the wind-solar power generation data obtained in the previous steps, the terminal inputs the wind-solar power generation power data sorted in chronological order and the weather environment data within the corresponding time period into the prediction model. Inside the model, by modeling the time series structure, it extracts time series feature information, such as power change trends, historical periodic fluctuations, etc., and constructs a high-dimensional feature vector representing the time evolution law of power. At the same time, the model takes the power data and meteorological data as joint inputs, models their mutual influence relationship through the attention mechanism, identifies the dependence degree of factors such as wind speed and irradiance on power output, and extracts the dynamic correlation features between multiple variables. After that, the model decodes by combining the above time series features and correlation features, generates the predicted value of wind-solar power generation power within the target prediction time period, and uses it as the input for subsequent energy system control optimization.

[0078] Optionally, the terminal can batch process the wind-solar power generation data in a sliding window manner to improve the prediction frequency and update accuracy; or it can set multiple prediction time granularities to form multi-level prediction results for different strategy selections.

[0079] Step S206, obtain the operating states of multiple energy units of the energy system.

[0080] Among them, the operating state can be a set of working parameters and status indicators of the energy unit at a certain moment, used to reflect the current capabilities and dispatchability of each energy unit.

[0081] In one embodiment, in the controlled energy system, the operating state of the current wind turbine can be characterized by the relationship between the power generation power of the wind turbine and air density, wind speed, etc.; the operating state of the battery can be characterized by the charge and discharge characteristics of the battery; the operating state of the hydrogen storage tank can be characterized by the hydrogen storage characteristics of the hydrogen storage tank; the operating state of the electrolyzer can be characterized by hydrogen production and electric power consumption; the operating state of the CHP combined heat and power unit can be characterized by the power generation power, natural gas consumption, and heat-electricity ratio of the CHP combined heat and power unit; the operating state of the gas boiler can be characterized by the relationship between the heat production power and natural gas consumption of the gas boiler; the operating state of the fuel cell can be characterized by the electric and heat output powers, conversion efficiency, and hydrogen consumption power of the fuel cell.

[0082] Step S208, based on the predicted wind-solar power generation power and the operating states, control multiple energy units under the control instructions in the control instruction set to determine the operating resources consumed by the energy system.

[0083] Among them, the control instruction set can be an optional control scheme generated for specific energy system states and predicted wind and solar power generation scenarios. The control instructions in the control instruction set include, but are not limited to, start-stop control, power distribution, electrolysis / discharge switching, etc. of multiple energy units. The operating resources can be the overall operating resource consumption of the energy system within the target prediction period after executing a certain set of control instructions. It can cover multiple dimensions, such as gas consumption, power loss, energy storage depreciation cost, energy waste, etc.

[0084] Exemplarily, based on the predicted wind and solar power generation obtained previously and the operating states of multiple energy units, the terminal sequentially selects control instructions from the control instruction set for simulation scheduling and calculates the operating resource consumption of the system under this instruction. First, the terminal uses the wind and solar power generation prediction result as the input load prediction basis, and together with the current operating states of each energy unit as the conditional input, sequentially loads several control instructions in the control instruction set. For each set of control instructions, the energy system is simulated and executed according to its indication, such as turning on or off a gas boiler, adjusting the discharge power of a battery, starting a fuel cell or an electrolyzer, etc. During this process, the resource usage of each energy unit is cumulatively calculated, including gas consumption, power utilization rate, energy storage service life loss, electrolysis / hydrogen storage efficiency loss, etc. The terminal combines the resource consumption of all energy units under this control instruction to obtain the operating resource index value corresponding to this set of control instructions. This process can be batch-executed for all control instruction schemes to form a resource evaluation result set for subsequent selection of the optimal control instruction.

[0085] Optionally, the terminal can set resource consumption weight coefficients (such as electricity price, gas price, depreciation factor) to adapt to the cost sensitivity requirements of different operating strategies.

[0086] Step S210, with the goal of minimizing the operating resources, determine the target control instructions for multiple energy units from the control instruction set.

[0087] Exemplarily, based on the determination of the operating resources, the terminal evaluates the resource cost of each set of control instructions in the control instruction set according to the objective function of minimizing the total operating resources, and compares their corresponding operating resource index values. Select the control instruction with the minimum resource consumption value as the target control instruction. This target control instruction will cover the operation contents of multiple energy units, such as a set of coordinated control operations like starting a CHP combined heat and power unit + turning off a gas boiler + discharging a battery + electrolyzing hydrogen.

[0088] Optionally, when there are multiple sets of control instructions with similar resource consumption, the terminal can also introduce a secondary priority strategy (such as equipment life, equipment start frequency, etc.) for further screening to ensure the smooth operation of the system and the health of the equipment.

[0089] In this embodiment, by obtaining the real-time power data of wind-solar power generation and environment-related data, and using a prediction model to deeply explore the temporal characteristics and internal correlations of these data, high-precision prediction of future wind-solar power generation power is achieved; on this basis, combined with the real-time operating states of multiple energy units in the energy system, dynamic optimization decisions are made with the goal of minimizing resources, reducing the curtailment rate of wind and light and the standby energy usage cost; at the same time, the dynamic scheduling strategy can reduce the frequent start-stop and overcharge / discharge of energy units, extend the service life of the equipment, and improve the real-time performance of the control of the energy system while ensuring the reliability of the system.

[0090] In an exemplary embodiment, based on the predicted wind-solar power generation power and operating state, multiple energy units are controlled under the control instructions in the control instruction set to determine the operating resources consumed by the energy system, including:

[0091] Determine the first operating state and the second operating state in the operating state; the first operating state is the operating state that occurs at the next moment of the second operating state; according to the first operating state and the second operating state, determine the working trend of the energy system; based on the predicted wind-solar power generation power and the working trend, control multiple energy units under the control instructions in the control instruction set to determine the operating resources consumed by the energy system.

[0092] Among them, the operating state is an influencing factor of the control instruction. The second operating state can be the system operating state at the current moment (t), and the first operating state can be the operating state at a certain time point (t + Δt) after the t moment. In an example, the second operating state S t can include the predicted power generation power of photovoltaic power generation, the predicted power generation power of wind turbines, electrical load, thermal load, hydrogen load, the state of charge of the battery, and the state of the hydrogen storage tank. At the moment t, it can be expressed by formula (1):

[0093] S t ={P WT (t), P PV (t), P eload (t), H load (t), P Hyload (t), S BA (t), Q HST (t), t} (1)

[0094] Among them, P WT (t) is the power generation power of the wind turbine at the moment t, in kW; P PV (t) is the power generation power of the photovoltaic at the moment t, in kW; P eload (t) is the electrical load demand at the moment t, in kW; H load (t) is the thermal load at the moment t, in kW; PHyload H(t) is the hydrogen load at time t, in kW; S BA B(t) is the state of the battery at time t; Q HST G(t) is the state of the hydrogen storage tank at time t.

[0095] In one embodiment, a wind turbine generator can be modeled, and the power generation power P WT (t) of the wind turbine at time t can be determined through formulas (2) and (3).

[0096]

[0097] where ρ is the air density, in kg / m 3 ; S is the area swept by the wind turbine, in m 2 ; V is the wind speed, in m / s; β is the pitch angle, in (°); λ i is the tip speed ratio.

[0098] In one embodiment, the battery is modeled, and the state-of-charge of the battery at time t is determined through formula (4).

[0099]

[0100] where P BA (t) is the charge and discharge power of the battery at time t, in kW. When the value of P BA (t) is greater than 0, the battery releases electrical energy to supplement the grid demand. When the value of P BA (t) is less than 0, the battery replenishes electrical energy; W BA is the total capacity of the battery, in kWh; η BA is the battery efficiency, which can be expressed by formula (5).

[0101]

[0102] where η ch is the charging efficiency of the battery; η dis is the discharging efficiency of the battery.

[0103] In one embodiment, the hydrogen storage tank is modeled, and the state-of-charge of the hydrogen storage tank at time t is determined through formulas (6) and (7).

[0104]

[0105] m HST (t) = m HST0 (t) + m hst，cha (t) - m hst，dis (t) / η hst，dis (7)

[0106] Among them, Q HST (t) is the state of the hydrogen storage tank at time t; m HST (t), m HST0 (t) are the hydrogen storage amount and the initial amount of the hydrogen storage tank at time t, with the unit of kg; W HsT is the rated capacity of the hydrogen storage tank, with the unit of kg; m hst,cha (t), m hst,dis (t) are the hydrogen charging and discharging amounts of the hydrogen storage tank at time t, with the unit of kg; η hst,cha , η hst,dis are the hydrogen charging and discharging efficiencies of the hydrogen storage tank.

[0107] In one embodiment, the electrolyzer is modeled, and the hydrogen molar flow rate of the electrolyzer at time t is determined by formulas (8) and (9).

[0108]

[0109] Among them, is the hydrogen molar flow rate of the electrolyzer at time t, with the unit of kmol / h; is the hydrogen production amount of the electrolyzer at time t, with the unit of kg; is the hydrogen production power of the electrolyzer within time t, with the unit of kW, and can be expressed by formula (10); L LHV is the lower calorific value of hydrogen, with a value of 0.067 kWh / mol; is the molar mass of hydrogen, with a value of 0.002 kg / mol.

[0110]

[0111] Among them, P EL (t) is the consumed electric power of the electrolyzer at time t, with the unit of kW.

[0112] In one embodiment, the gas turbine of the CHP combined heat and power unit is modeled, and the power generation power of the CHP combined heat and power unit at time t is determined by formula (11).

[0113] P CHP (t) = V CHP (t)H NG η CHP (11)

[0114] Among them, P CHP (t) is the electric power output by the CHP combined heat and power unit at time t, with the unit of kW; V CHP (t) is the natural gas consumption of the CHP combined heat and power unit at time t, with the unit of m 3 ; H NG is the calorific value of natural gas, with the unit of kWh / mol; η CHPis the power generation efficiency of the CHP unit. The heat - to - power ratio is the ratio of the actual output heat power to the electric power of the CHP unit, which can be expressed by formula (12).

[0115]

[0116] Among them, b is the heat - to - power ratio of the CHP unit; H CHP (t) is the heat supply power of the CHP unit at time t, with the unit of kW.

[0117] In one embodiment, a gas boiler is modeled, and the heat generation power of the gas boiler at time t is determined by formula (13).

[0118] H GB (t)=V GB (t)H NG η GN (13)

[0119] Among them, V GB (t) is the natural gas consumption of the gas boiler at time t, with the unit of m 3 ; η GB is the heat generation efficiency of the gas boiler.

[0120] In one embodiment, a fuel cell is modeled. The fuel cell is an important coupling element of the above - mentioned energy system, which can convert the chemical energy in hydrogen into electric energy to meet the needs of the power system imbalance. The state of the fuel cell at time t can be determined by formulas (14) - (17).

[0121]

[0122] Among them, P FC (t), H FC (t), η FC , and are the electric and heat output powers (kW), conversion efficiency, hydrogen consumption power (kW), and heat consumption power (kW) of the fuel cell; is the hydrogen molar flow rate of the fuel cell, with the unit of mol; is the hydrogen consumption of the fuel cell, with the unit of kg.

[0123] Among them, the working trend can be the change direction and change amplitude of the energy system between adjacent operating states. The feedback (reward or punishment) of the energy system generated by the control instruction can be determined through the working trend, so as to adjust the next control instruction.

[0124] Exemplarily, the terminal obtains the operating status information at two consecutive time points, denoted as the first operating status and the second operating status. Compare them item by item to determine the current working trend of the system. For example, if the heat load continuously increases and the load of the CHP (Combined Heat and Power) unit is close to full load, it is considered that there is a tight trend in the heating load, that is, the current control instruction has a punitive impact on the energy system; if the output of wind and light slightly decreases, but the state of charge of the battery continuously rises, it indicates that the current trend of renewable energy consumption is enhanced, that is, the current control instruction has a rewarding impact on the energy system.

[0125] The terminal uses this working trend and the predicted wind power generation power as inputs, loads the control instructions in the control instruction set, simulates and executes them respectively, and calculates the resource consumption under the influence of the working trend. For example, when the working trend indicates that the current control instruction has a punitive impact on the energy system, a control instruction opposite to the previous control instruction will be selected from the control instruction set to control the energy system, so that the control of the energy system approaches a favorable direction, and the response speed is improved, enabling rapid control of the energy system. For example, when the wind power generation efficiency drops suddenly, energy storage switching can be completed in a short time, the power generation efficiency of the electrolyzer and the gas supply unit of the CHP unit can be improved, and the waste energy loss is reduced.

[0126] Output the operating resources consumed under each group of control instructions, providing an evaluation basis guided by the trend for the subsequent screening of target control instructions.

[0127] Specifically, as Figure 3 shown, the terminal can train an agent to select control instructions through the PPO (Proximal Policy Optimization, a policy gradient reinforcement learning algorithm) reinforcement learning framework to achieve coordinated control of multiple units of the energy system. In each control cycle, the terminal constructs a state vector S t , and calculates the working trend. Based on the state vector S t and the working trend, a set of action control instructions is selected from the control instruction set A t . The control instruction set can be represented by formula (18).

[0128] A t ={P CHP (t),P EL (t)} (18)

[0129] Where, P CHP (t),P EL(t) are the power generation powers of the CHP combined heat and power unit and the gas combined heat and power unit, and the power generation power of the electrolyzer respectively, which are used to indicate the control actions on the CHP combined heat and power unit and the electrolyzer, so that the power generation powers of the CHP combined heat and power unit and the gas combined heat and power unit and the power generation power of the electrolyzer are changed, and the operating state of the entire energy system is changed.

[0130] To make the execution of the determined control instruction beneficial to the energy system and to accelerate learning at the same time, the terminal will determine the impact (reward or punishment) of the control instruction on the energy system according to the working trend, so as to determine the control instruction. The impact of the control instruction on the energy system can be represented by the reward function R t as shown in formula (19).

[0131] R t = -γ R (C BA (t) + C gas (t) + C m (t) + C p (t) - δ R (D waste (t) + D gas (t) + D Rt (t)) (19)

[0132] Where γ R , δ R are the hyperparameters of the reward function cost and punishment; D waste (t) is the energy waste penalty; D gas (t) is the penalty for exceeding the limit of natural gas purchase; D Rt (t) is the penalty for the equipment operation exceeding the safe range.

[0133] In one embodiment, if after executing the previous control instruction, the target trend indicates that the operating resources consumed by the energy system are reduced, it is determined that the previous control instruction is beneficial to the energy system, a reward; if after executing the previous control instruction, the target trend indicates that in the energy system, the energy storage upper limits of the energy storage devices (hydrogen storage tanks and storage batteries) are exceeded, that is, the curtailment rate of wind and solar power is high, or the working efficiency of the CHP combined heat and power unit and the gas combined heat and power unit is low, that is, the normal energy supply to the load cannot be carried out, or the working loads of each energy unit are too high, that is, the energy unit has exceeded the safe working range, it is determined that the previous control instruction is a punishment for the energy system. The terminal can determine the control instruction for the next moment according to the impact of the control instruction indicated by the target trend on the energy system.

[0134] In this embodiment, by calculating the differences between the first and second operating states, the resource evaluation is provided with a sense of time direction, realizing dynamic compensation for prediction errors and avoiding the problem of policy lag caused by static scheduling. Through trend-driven guidance, control instructions are used to more rationally utilize energy units, especially reducing the dependence on traditional energy when renewable energy fluctuations are significant. Thus, under the premise of ensuring operating safety and meeting the load demand, wind and solar power curtailment is reduced, thereby improving the utilization rate of energy.

[0135] In an exemplary embodiment, the operating resources include the operating resources consumed by each energy unit. With the goal of minimizing the operating resources, the target control instructions for multiple energy units are determined from the control instruction set, including:

[0136] Determine the energy units according to the operating resources consumed by each energy unit; obtain the operating parameters of the energy units under the control instructions, and compare the operating parameters with the preset parameter thresholds to obtain a comparison result; the comparison result is used to indicate the control priority of the energy units; based on the control priority, with the goal of minimizing the operating resources, determine the target control instructions for multiple energy units from the control instruction set; the target control instructions include the opening and closing of the energy units.

[0137] Among them, the energy unit is the unit with a large consumption or the most influential on the resource optimization goal in the energy system. The operating parameters can be the operating indicators of the energy unit under the current control instructions, such as output power, equipment load rate, energy storage state, etc., reflecting its operating state and performance boundaries.

[0138] Among them, the preset threshold can be the boundary limit that satisfies the power balance constraint conditions of the energy system, used to reflect the safety and stability of each energy unit, and can include safety capacity, equipment limit, energy supply boundary, etc.

[0139] In one embodiment, the power balance constraint condition can be the electric power balance constraint, which is represented by formula (20).

[0140] P e (t)=P CHP (t)P WT (t)+P PV (t)+P BA (t)+P FC (t)-P EL (t) (20)

[0141] Among them, P e (t) is the electrical load demand at time t, with the unit of kW; P WT (t) is the power generation of the wind turbine at time t, with the unit of kW; P PV (t) is the power generation of the photovoltaic at time t, with the unit of kW.

[0142] In one embodiment, the power balance constraint condition can be a hydrogen energy balance constraint, which is represented by formula (21).

[0143]

[0144] where m HST (t) is the hydrogen charging amount in the hydrogen storage tank at time t, with the unit of kg; m Hyload (t) is the hydrogen load at time t, with the unit of kg.

[0145] In one embodiment, the pre-power balance constraint condition can be a thermal power balance constraint, which is represented by formula (22).

[0146] H GB (t)+H CHP (t)+H FC (t)=H load (t) (22)

[0147] where H FC (t) is the heat generation power of the hydrogen fuel cell at time t, with the unit of kW.

[0148] In one embodiment, the preset parameter threshold can be the safety constraint of energy conversion devices (CHP combined heat and power unit, electrolyzer, gas boiler), which is represented by formulas (23) and (24). The capacities of the battery and the hydrogen storage tank have limited ranges to ensure the normal operation of the devices, which can be expressed as:

[0149] S BAmin (t)≤S BA (t)≤S BAmax (t) (23)

[0150] Q HSTmin (t)≤Q HST (t)≤Q HSTmax (t) (24)

[0151] where S(t) BAmin 、S(t) BAmax are the lower limit and upper limit of the state of charge of the battery. In this embodiment, at the beginning of the time range, the battery is charged to 50%; Q HSTmin (t)、Q HSTmax (t) are the lower limit and upper limit of the state of the hydrogen storage tank. In this embodiment, at the beginning of the time range, the initial capacity of the hydrogen storage tank is 50%.

[0152] In one embodiment, the preset threshold can be the safety constraint of energy conversion devices (CHP combined heat and power unit, electrolyzer, gas boiler), which is represented by formulas (25)-(28).

[0153]

[0154] Among them, they are the power constraints of the CHP combined heat and power unit's electric and heat power safety constraints, the upper and lower limits of the electrolyzer's power consumption, and the heat production power of the gas boiler, respectively.

[0155] In one embodiment, the preset threshold can be the energy supply (natural gas supply) safety constraint, which is expressed by formulas (29)-(30).

[0156] V gh,min (t) ≤ V gh (t) ≤ V gh，max (t) (29)

[0157] V gh (t) = V GB (t) + V CHP (t) (30)

[0158] Among them, V gh (t) is the total amount of natural gas purchased from the natural gas network, with the unit of m 3 ; V gh，max (t), V gh，min (t) are the upper and lower limits of natural gas purchase, with the unit of m 3 .

[0159] Among them, the operating resources can be the sum of the resource consumptions of all energy units in the energy system during a certain period, and can include gas usage, electricity usage cost, equipment depreciation, electrolysis power consumption, curtailment of wind and solar power costs, etc., which are used to measure the overall operating cost and are the direct indicators for control optimization.

[0160] In one embodiment, to reduce the system operating cost and improve the energy utilization rate of the system, the stored electricity of the energy storage device battery and the stored hydrogen volume of the hydrogen storage tank should be ensured within a safe range, and the battery charging and discharging should not exceed its safety limit. The operating resources are characterized as natural gas cost (gas purchase cost), battery depreciation cost, equipment maintenance cost, and curtailment of wind and solar power costs, which are expressed by formula (31).

[0161]

[0162] Among them, C BA (t) is the battery depreciation cost, with the unit of yuan; C has (t) is the natural gas purchase cost, with the unit of yuan; C m (t) is the system operating cost, with the unit of yuan; C p (t) is the curtailment of wind and solar power cost, with the unit of yuan. Specifically, it can be expressed by formulas (32)-(35).

[0163]

[0164] Among them, η BA is the battery depreciation coefficient; α g is the natural gas price, with the unit of yuan / m 3 ; V g (t) is the natural gas consumption, with the unit of m 3 ; η is the conversion coefficient; P WT,r (t), P PV,r (t) are the available powers of wind and light, with the unit of kw; i is the i-th device.

[0165] Exemplarily, under a given set of control instructions, the terminal calculates the operating resource consumption of each energy unit, such as the natural gas consumption of a CHP combined heat and power unit, the power consumption of an electrolyzer, the battery discharge loss, the curtailment losses of wind turbines and photovoltaics, etc.

[0166] Since there is a power exchange relationship among the energy units in the energy system, after the terminal obtains the operating parameters of each energy unit, it compares the operating parameters with the preset parameter thresholds constructed based on the power balance constraint of the energy system to obtain a comparison result. In one embodiment, under the given control instructions, if the load state of the battery is close to the lower limit of the preset parameter threshold, the control priority of the battery can be increased to optimize the charging and discharging timing, reduce the number of charge and discharge cycles of the battery, and extend its life; if the operation of the energy conversion device is close to the upper limit of the preset parameter threshold power, the control priority of the energy conversion device can be reduced to lower the start-stop frequency of the energy conversion device, reduce the maintenance cost, and extend the service life of the energy conversion device. It can realize the storage of hydrogen produced by surplus photovoltaics during the day and the power generation of hydrogen energy during the low wind power period at night, which will greatly improve the consumption capacity of renewable energy.

[0167] The terminal maps the comparison results corresponding to the operating parameters of each energy unit to the control priorities of the corresponding energy units, and filters out the control instructions that meet the control priorities of the energy system and the minimum operating resources in the set of control instructions as the target control instructions to realize the dynamic switching of the hydrogen production and power generation modes based on the real-time capacity threshold.

[0168] In this embodiment, by judging the device operating parameters and the preset parameter thresholds, the dynamic adjustment of the safety constraints of the policy selection is realized, effectively preventing the device from operating out of bounds or starting and stopping frequently. Based on the priority allocation, the logical closed-loop between the resource scheduling order and the generation of control instructions is realized, improving the robustness and response efficiency of the policy execution, optimizing the control instructions of the energy system, and breaking through the collaborative bottleneck of the efficiency and life of the energy unit.

[0169] In an exemplary embodiment, the temporal features of wind and solar power generation data and the correlation between current wind and solar power generation data and weather environment data are extracted through a prediction model. According to the temporal features and the correlation, the wind and solar power generation is predicted to obtain the predicted wind and solar power generation, including:

[0170] The wind and solar power generation data is input into an encoder network to extract an encoded result representing the temporal features; the weather forecast data of the weather environment data for a preset time period is obtained, and the encoded features composed of the encoded result and the weather forecast data are input into a decoder network. The intermediate layer of the decoder network determines the intermediate features corresponding to the encoded features, and by fusing the encoded features and the intermediate features, a fused feature is obtained; the fused feature indicates the correlation between the current wind and solar power generation data and the weather environment data; the fused feature is input into the fully connected layer of the decoder network to output the predicted wind and solar power generation.

[0171] Exemplarily, as Figure 4 shown, the prediction model adopts an improved LSTM model, and the prediction model includes an encoder network and a decoder network. In one embodiment, the wind power of the wind and solar power generation is predicted. The encoder network can adopt a multi-layer recursive network. In one embodiment, a two-layer stacked LSTM structure is adopted, that is, the historical wind power data and the corresponding time encoding are first input into the first-layer LSTM unit for processing. This layer can normalize the output of the encoder network and identify the data fluctuations and time interval characteristics in the short term; its output is then transmitted to the second-layer LSTM unit to further extract the cross-periodic trend or the power generation change pattern on a longer time scale. The two-layer structure superimposition enhances the model's expression ability for time series data, and the output is the encoded result.

[0172] In the decoder network, an LSTM unit is included as the intermediate layer for processing the input of the decoder network, and on this basis, a multi-source input fusion mechanism and a residual connection mechanism are introduced. The input of the decoder not only includes the encoded structure output by the encoder, but also fuses the weather forecast data for the future time period and the corresponding time variable encoding. After being processed by the LSTM unit of the intermediate layer, the intermediate features are obtained, so that the prediction model can perceive the upcoming meteorological changes when generating the output. After the processing of the LSTM unit, the model introduces a residual connection path, that is, the encoded result (historical time series representation) output by the encoder is directly connected to the intermediate features extracted by the intermediate layer of the decoder to obtain the fused feature. This structure helps the model retain the original historical information, indicates the correlation between the current wind and solar power generation data and the weather environment data, and at the same time embeds the adjustment effect generated by the weather change. The fused feature is fed into the fully connected layer at the end of the decoder to map the fused feature representation to the predicted value of the specific wind power. In one embodiment, for the optical power, the same above prediction model as that for the wind power can be adopted to output the predicted optical power.

[0173] In this embodiment, the encoder is used to extract the time series characteristics of historical data, so that the system can deeply understand the time evolution characteristics of the power generation process, thereby avoiding the prediction deviation caused by simple linear fitting, especially in the scene where there are fluctuations in wind speed, light, etc., the model has a stronger capture ability. Combined with residual connection, the encoder output and historical information are fused, and the numerical forecast information of the time period is introduced to improve the accuracy of the prediction.

[0174] In an exemplary embodiment, the prediction model is obtained through the following steps: acquiring historical data of wind and solar power generation, and classifying the historical data of wind and solar power generation according to preset classification conditions to obtain multiple categories of historical data of wind and solar power generation; for each category of historical data of wind and solar power generation, performing model training according to the category of historical data of wind and solar power generation to obtain a prediction sub-model corresponding to the category of historical data of wind and solar power generation; integrating multiple prediction sub-models to obtain a prediction model.

[0175] Among them, the classification conditions are applicable to the criteria for dividing the historical data of wind and solar power generation into multiple subsets, so as to improve the model's ability to express the rules under different data distributions. The classification conditions of the historical data of wind and solar power generation can be determined by clustering algorithms.

[0176] Exemplarily, when constructing a wind and solar power generation prediction model, the terminal adopts a clustering-based data partitioning and sub-model construction mechanism to better characterize the power generation characteristics of wind and solar power generation as the environment changes. Specifically, DPC (Density Peak Clustering), a clustering method based on density peak, is used. It does not require the number of clusters to be specified in advance and is suitable for data of complex shapes. The historical data of wind and solar power generation is divided into several data clusters, each of which represents a sub-scenario or output mode with similar power generation behavior characteristics. After clustering is completed, the terminal trains a prediction sub-model separately based on each clustering result. Each sub-model can adopt the same or different structures. The specific model structure is such as an encoder-decoder neural network, which supports the joint modeling of time series data and weather variables. During the training process, each sub-model only receives the data of the cluster cluster to which it belongs for training, fully explores the power generation behavior laws of this type of data, and avoids the problem of reduced prediction accuracy of the global model due to excessive differences in data distribution. After all sub-models are trained, the terminal integrates them uniformly to form a prediction system with a combination of multiple sub-models. This integration mechanism can support dynamic scheduling, that is, automatically matching the most appropriate sub-model for prediction according to the input data characteristics during the inference stage, and can also use the model fusion mechanism to perform weighted combination of the outputs of multiple models.

[0177] After obtaining wind and solar power generation data, it also includes:

[0178] Wind and solar power generation data belonging to different categories are determined, and the wind and solar power generation power corresponding to each category of wind and solar power generation data is predicted according to the prediction sub-models corresponding to the wind and solar power generation data of different categories to obtain the predicted wind and solar power generation power.

[0179] Specifically, when actually making predictions, after receiving new wind and solar power generation data, the terminal compares the data with the historical cluster centers for similarity, thereby determining the historical cluster cluster to which the data is most likely to belong, and further determining the category of the wind and solar power generation data. After determining the category, the terminal calls the corresponding prediction sub-model, inputs the current data, and outputs the prediction results. If the input data spans multiple cluster boundaries, it can also support parallel prediction of multiple sub-models and fuse their results to improve the robustness of the prediction.

[0180] In this embodiment, by introducing the classification mechanism and sub-model training strategy in the model construction stage, it not only ensures that the prediction model still has good generalization ability when facing nonlinear and violently fluctuating wind and solar power generation data, but also improves the prediction stability and accuracy of the system under extreme climate or special load scenarios. It integrates the robustness enhancement mechanism driven by multimodal meteorological data to solve the difficult problem of balancing edge computing power and accuracy.

[0181] In an exemplary embodiment, model training is performed according to the historical data of wind and solar power generation of the category to obtain a prediction sub-model corresponding to the historical data of wind and solar power generation of the category, including:

[0182] Input the historical data of wind and solar power generation into the prediction sub-model to be trained, and the prediction sub-model to be trained determines the hidden state information of the corresponding time series characteristics in the historical data of wind and solar power generation, and removes the weighted processing results of the hidden state information from the historical data of wind and solar power generation to obtain intermediate training data; determine the degree of change of the weather environment corresponding to the weather environment data of the historical data of wind and solar power generation, and use the intermediate training data and the degree of change of the weather environment as target training data; the prediction sub-model to be trained determines the training predicted wind and solar power generation according to the target training data; according to the difference between the training predicted wind and solar power generation and the actual wind and solar power generation of the historical data of wind and solar power generation, adjust the parameters of the prediction sub-model, and update the hidden state information and the weights corresponding to the hidden state information; return to execute the step of removing the weighted processing results of the hidden state information from the historical data of wind and solar power generation to obtain the intermediate training data, until the preset training conditions are met and the prediction sub-model of the corresponding category is obtained.

[0183] Specifically, the terminal trains the corresponding prediction sub-model for the historical data of wind and solar power generation under a certain clustering category. In the initial stage of training, the historical data of wind and solar power generation of this category is input into the model to be trained, and the model is based on the long short-term memory structure.

[0184] The hidden state of the previous moment and the input of the current moment are jointly processed through the forget gate to generate a forget vector. The forget gate can be implemented by formula (36). The forget gate can determine the hidden state information corresponding to the time series features in the historical data of wind and solar power generation, and by adjusting the weight initialization of the forget gate, the weighted processing result of the hidden state information is removed from the historical data of wind and solar power generation to obtain intermediate training data, so as to enhance the model's ability to retain long-term correlation features, thereby avoiding the mistaken deletion of key trend information due to short-term mutations under conditions such as cloudy weather.

[0185] f t =σ(W f [h t-1 , x t ]+b f ) (36)

[0186] Among them, f t is the output of the forget gate, σ is the sigmoid activation function, W t is the weight matrix of the forget gate, b f is the bias term, h t-1 is the hidden state at the previous moment, x t is the input at the current moment.

[0187] On this basis, the model uses the input gate to determine which content in the current input should be written into the memory unit. The input gate can be implemented using formulas (37) and (38). The input gate introduces the degree of change in weather conditions such as solar irradiance as an input gate adjustment factor. The weight of the input gate output controls the proportion of new information entering, ensuring that the model is more inclined to capture data dimensions with drastic feature changes in the input and high correlation with the output.

[0188] i t =σ(W i [h t-1 , x t ], b i ) (37)

[0189]

[0190] Among them, i t is the output of the input gate, is the candidate memory cell state.

[0191] The results of the forget gate and the input gate are combined through formula (39) and used as the final target training data to participate in the forward training process of the model.

[0192]

[0193] Among them, C tis the memory cell state at the current moment.

[0194] In the output stage, the model adjusts which parts of the current memory state are used to generate the final hidden state through the output gate, and this hidden state directly affects the prediction of the power generation. Among them, the output gate can be implemented by formulas (40) and (41). The output gate not only calculates the result based on the current state but also combines the weather category label to which the sample belongs, dynamically adjusting the sensitivity of the output path to the sudden weather characteristics and improving the response speed in non-stable scenarios. The model outputs the predicted wind and solar power generation values at this stage and calculates the error with the actual wind and solar power generation values in the historical data. The error feedback is not only used for regular parameter updates but also further acts on the weight adjustment process in the gating mechanism, that is, the model redistributes the hidden state weights according to the prediction deviation, dynamically strengthening the contributing input paths while weakening or eliminating the invalid input paths.

[0195] O t = σ(W o [h t-1 , x t +b o ) (40)

[0196] h t = O t tanh(C t ) (41)

[0197] Among them, O t is the output of the output gate, and h t is the hidden state at the current moment.

[0198] This training process is carried out in a closed-loop manner: after each prediction and error feedback, the model will re-evaluate the distribution of the hidden state, reconstruct the intermediate training data, and perform iterative training by combining the weather environment information again. Optionally, the cosine annealing algorithm is used to dynamically adjust the learning rate to avoid falling into local optima. Optionally, according to the category of the prediction sub-model being trained currently, it is determined whether it is a high-fluctuation category, and a higher weight is set for the prediction error of the high-fluctuation category to optimize the model's prediction ability for extreme weather.

[0199] Until the model meets the set training convergence condition for the current category, the terminal determines that the training of the prediction sub-model corresponding to this category is completed.

[0200] In this embodiment, the entire training process iterates multiple rounds in a closed loop of prediction, feedback, and correction. Each round includes hidden state update, feature elimination, intermediate sample reconstruction, and parameter weight adjustment until the model meets the set training stop condition, and finally forms a prediction sub-model corresponding to the wind-solar data of this category. By adjusting the forgetting gate weight initialization strategy, the memory ability for long-term historical power sequences is strengthened, and the prediction lag caused by sudden changes in light under cloudy weather is avoided. Features such as the temporal gradient features of solar irradiance, which reflect the degree of weather environment changes, are introduced into the input gate. At the same time, the output gate combines the class labels of the prediction sub-model, improving the stability and prediction accuracy of the model under complex climate conditions.

[0201] In an exemplary embodiment, after obtaining the wind-solar power generation data, the wind-solar power generation data is preprocessed.

[0202] Specifically, for the common outlier and fault data problems in wind power generation data, the terminal introduces an anomaly detection mechanism based on One-Class SVM (One-Class Support Vector Machine). The core decision-making process of the OneSVM algorithm can be expressed by formula (42).

[0203]

[0204] where w is the weight vector of the decision function; ρ is the intercept of the decision function; ξ i is the slack variable; n is the number of training samples; v is a hyperparameter used to control the proportion of outlier samples.

[0205] This algorithm uses historical wind power samples to construct a boundary, and combines with the RBF kernel function, that is, the radial basis function kernel, to realize the non-linear mapping from the input space to the high-dimensional feature space, enabling the model to identify outlier samples far from the main data distribution without relying on outlier labels. The RBF kernel function can be expressed by formula (43).

[0206] K(x, x′) = exp(-γ||x - x′|| 2 ) (43)

[0207] where x and x′ are input sample points, γ is a parameter that controls the steepness of the function curve, and ||x - x′|| 2 represents the square of the Euclidean distance between sample points.

[0208] By reasonably setting the slack variable and the hyperparameter for anomaly rate control, OneSVM can effectively identify and remove the power mutation segments in the wind power sequence caused by wind turbine failures, curtailment scheduling, etc., thus preventing the model from mistakenly incorporating atypical patterns into the learning process. In this process, the RBF kernel function enhances the accuracy of the model in depicting the boundaries of anomaly points through its non-linear mapping ability of Euclidean distance. This method effectively guarantees the convergence speed and prediction stability of subsequent LSTM-based models for wind power time series.

[0209] For photovoltaic power generation data, the terminal first performs feature selection, screening out five features highly correlated with photovoltaic output from meteorological and power variables, including air temperature, cloud opacity, solar irradiance, relative humidity, and historical output power. Among them, solar irradiance has the highest correlation with photovoltaic output, with a correlation coefficient exceeding 0.95. To unify the feature dimensions and improve the consistency of model inputs, the terminal performs min-max normalization on the above feature data, mapping all features to the [0, 1] interval to avoid unbalanced network training caused by inconsistent dimensions. The normalization step can be expressed by formula (44). In addition, to further compress the feature dimensions and extract non-linear distribution structure information, the terminal uses the t-SNE dimensionality reduction method to compress the five-dimensional features into a two-dimensional representation. This dimensionality reduction strategy preserves the neighborhood structure of the data in the high-dimensional space, helping the subsequent prediction model to more effectively identify the typical patterns and change trajectories of photovoltaic output.

[0210]

[0211] Among them, z i is the data after min-max normalization; y i is the original data; y min is the minimum value of the original data features; y max is the maximum value of the original data features.

[0212] Through the above steps, the terminal has completed the cleaning, structural optimization, and expression reconstruction of wind power and photovoltaic data, improving the usability and information density of the input data, and providing stable and high-quality input support for the subsequent power prediction model based on the deep time series structure. This preprocessing strategy exhibits good robustness and generalization performance in scenarios of extreme weather fluctuations and high-frequency power disturbances.

[0213] In an exemplary embodiment, an energy system typically includes energy units such as wind turbines, photovoltaic modules, energy conversion units (CHP combined heat and power units, gas boilers), P2G devices (electrolyzers), and batteries / hydrogen storage tanks. To determine the control instructions for the energy system, the steps can be as Figure 5 shown.

[0214] Step S501, acquisition and preprocessing of wind and solar power generation data. Real-time output power data of wind turbines and photovoltaic modules under different meteorological conditions and weather environment data such as wind speed, irradiance, temperature, humidity, and cloud cover are acquired to comprehensively collect wind and solar power generation data. In order to ensure the validity of input data, it is necessary to eliminate abnormal values or faulty data, identify and filter out samples that deviate too much from the main data distribution through methods such as single-class support vector machines, and then perform maximum and minimum normalization or kernel function-based dimensionality reduction processing on the remaining power generation data and meteorological data, so as to obtain stable learning features in the subsequent model building stage.

[0215] Step S502, construction and training of wind and solar power generation prediction model. A clustering-based strategy is used to divide the historical wind and solar power generation data according to distribution characteristics. Each sub-category of data corresponds to a typical weather pattern, and a prediction sub-model is trained for each sub-category. A network structure with a gating mechanism such as LSTM can be used inside the sub-model, and multi-dimensional correlation extraction and fusion of historical time series features, meteorological information, and current power status are performed according to the input gate, forget gate, and output gate. The future weather forecast data is combined with the historical power law through residual connection and attention mechanism to generate the prediction result of wind and solar power. After completing the training of all sub-category data, these sub-models are integrated to form a comprehensive wind and solar power prediction model for multiple meteorological scenarios.

[0216] Step S503, using the prediction model to predict the wind and solar power generation. During the official operation, after the system collects wind and solar power generation data in real time, it determines which category the current data belongs to based on its similarity with the divided cluster center, calls the corresponding sub-model to predict the wind and solar power generation in the subsequent period, and further corrects the power fluctuation trend in the short term in combination with weather forecast information to determine the wind and solar power generation.

[0217] Step S504, mathematical modeling is performed on each energy unit of the energy system to determine the operating status of each energy unit. The real-time operating status of each energy unit in the electric-hydrogen coupling system is obtained, including the state of charge and charging and discharging power of the battery, the hydrogen storage capacity and charging and discharging efficiency of the hydrogen storage tank, the power consumption and hydrogen production rate of the electrolyzer, the natural gas consumption and electricity-to-heat ratio of the CHP combined heat and power unit gas unit, the heating power of the gas boiler, and the hydrogen consumption and output power of the fuel cell, etc., and this information is uniformly recorded as the current operating status of the system.

[0218] Step S505: Determine the impact of the control instruction on the energy system according to the adjacent operating states, determine a new control instruction in the control instruction set to control the energy system, and calculate the operating resources consumed by the energy system. To achieve optimal control of the overall system energy consumption, a set of control instruction sets need to be defined, covering operations such as start / stop instructions for each device, power distribution, or mode switching. Then, the predicted wind and solar power generation and the current operating state are used together as control inputs, and different instructions in the control instruction set are simulated one by one to evaluate the resource consumption and feasibility after executing these instructions. To better guide the selection of control instructions, the operating states at adjacent times can be compared to judge the working trend of the energy system under the current control instruction. If the working trend is beneficial to the energy system, it means that the current control is conducive to reducing energy consumption and can be appropriately strengthened. If situations such as energy storage overlimit, excessive natural gas usage, or excessive curtailment rates of wind and solar power occur, the next instruction should be corrected.

[0219] Step S506: Under the constraint conditions of the energy units of the energy system, with the goal of minimizing operating resources, determine the target control instruction. The resource consumption of the energy system includes both explicit costs such as natural gas usage, power loss, and curtailment losses of wind and solar power, and potential depreciation costs caused by overcharging / discharging or frequent start / stop of equipment. On the other hand, the energy storage unit and the hydrogen energy conversion unit are restricted by safety thresholds at different power levels, and it is necessary to ensure that the state of charge of the battery, the hydrogen storage capacity of the hydrogen storage tank, the operating power of the CHP combined heat and power unit and the gas boiler, etc. are all within the specified upper and lower limits. If it is found that a certain control instruction will cause key equipment to operate beyond the limit or lead to a significant increase in curtailment of wind and solar power, a penalty should be imposed on it and its priority in subsequent selection should be reduced. The operating states at adjacent times can also be used to infer the working trend of the entire system. If the system energy consumption decreases after executing similar instructions twice in a row, it means that this instruction is relatively advisable in the current situation. If it has a negative impact, the strategy should be adjusted as soon as possible or other instructions should be switched to continuously optimize the system energy consumption while ensuring the supply of load demand.

[0220] In this embodiment, through a multi-level collaborative process of "data-driven prediction - dynamic optimization decision - real-time closed-loop control". By providing high-precision power prediction to quantify the future output trend and uncertainty, the energy management system generates a collaborative strategy for multi-energy devices based on the predicted data, aiming to optimize energy storage charging / discharging, hydrogen production / generation switching, and combined heat and power supply. This process deeply integrates the predictability of prediction and the flexibility of management, and finally realizes the multi-objective collaboration of efficient consumption of renewable energy, minimization of system operation costs, and grid security and stability.

[0221] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are sequentially shown according to the indications of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this document, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.

[0222] Based on the same inventive concept, an embodiment of the present application further provides a control device for an energy system considering wind and solar power generation for implementing the above-mentioned energy system control method considering wind and solar power generation. The implementation solutions provided by this device to solve problems are similar to those described in the above method. Therefore, the specific limitations in one or more embodiments of the control device for an energy system considering wind and solar power generation provided below can refer to the limitations on the energy system control method considering wind and solar power generation in the above text, and will not be repeated here.

[0223] In an exemplary embodiment, as Figure 6 shown, a control device for an energy system considering wind and solar power generation is provided, including: a wind and solar power generation data acquisition module 610, a wind and solar power generation prediction module 620, an energy unit operation state acquisition module 630, an operation resource determination module 640, and a target control instruction determination module 650, where:

[0224] The wind and solar power generation data acquisition module 610 is used to acquire wind and solar power generation data; the wind and solar power generation data includes current wind and solar power generation power data and weather environment data;

[0225] The wind and solar power generation prediction module 620 is used to extract the time series characteristics of the wind and solar power generation data and the correlation between the current wind and solar power generation power data and the weather environment data through a prediction model, and predict the wind and solar power generation power according to the time series characteristics and the correlation to obtain the predicted wind and solar power generation power;

[0226] The energy unit operation state acquisition module 630 is used to acquire the operation states of multiple energy units of the energy system;

[0227] The operation resource determination module 640 is used to control the multiple energy units under the control instructions in the control instruction set based on the predicted wind and solar power generation power and the operation states, and determine the operation resources consumed by the energy system;

[0228] A target control instruction determination module 650, configured to determine, with the goal of minimizing the operating resources, target control instructions for the multiple energy units from the set of control instructions.

[0229] In one embodiment, the operating resource determination module 640 is further configured to determine a first operating state and a second operating state in the operating state; the first operating state is the operating state that occurs at the next moment of the second operating state;

[0230] Determine the working trend of the energy system according to the first operating state and the second operating state;

[0231] Based on the predicted wind and solar power generation and the working trend, control the multiple energy units under the control instructions in the set of control instructions, and determine the operating resources consumed by the energy system.

[0232] In one embodiment, the target control instruction determination module 650 is further configured to obtain the operating parameters of the energy unit under the control instruction, and compare the operating parameters with a preset parameter threshold to obtain a comparison result; the comparison result is used to indicate the control priority of the energy unit;

[0233] Based on the control priority, with the goal of minimizing the operating resources, determine target control instructions for the multiple energy units from the set of control instructions; the target control instructions include the start and stop of the energy unit.

[0234] In one embodiment, the prediction model includes an encoder network and a decoder network; the wind and solar power generation prediction module 620 is further configured to input the wind and solar power generation data into the encoder network to extract an encoding result representing the timing characteristics;

[0235] Obtain weather forecast data of the weather environment data for a preset time period, and input the encoding features composed of the encoding result and the weather forecast data into the decoder network. The intermediate layer of the decoder network determines intermediate features corresponding to the encoding features, and by fusing the encoding features and the intermediate features, a fusion feature is obtained; the fusion feature indicates the correlation between the current wind and solar power generation data and the weather environment data; input the fusion feature into the fully connected layer of the decoder network to output the predicted wind and solar power generation.

[0236] In one embodiment, the wind and solar power generation prediction module 620 is further configured to obtain wind and solar power generation historical data, and classify the wind and solar power generation historical data according to preset classification conditions to obtain wind and solar power generation historical data of multiple categories;

[0237] For the historical wind and solar power generation data of each of the said categories, model training is performed based on the historical wind and solar power generation data of the category to obtain a prediction sub-model corresponding to the historical wind and solar power generation data of the category;

[0238] Integrate multiple said prediction sub-models to obtain the prediction model;

[0239] After obtaining the wind and solar power generation data, it further includes:

[0240] Determine the wind and solar power generation data belonging to different categories, so as to predict the wind and solar power generation power corresponding to the wind and solar power generation data of each category according to the prediction sub-models corresponding to the wind and solar power generation data of the different categories, and obtain the predicted wind and solar power generation power.

[0241] In one embodiment, the wind and solar power generation power prediction module 620 inputs the historical wind and solar power generation data into the prediction sub-model to be trained. The prediction sub-model to be trained determines the hidden state information corresponding to the time series features in the historical wind and solar power generation data, and removes the weighted processing result of the hidden state information from the historical wind and solar power generation data to obtain intermediate training data;

[0242] Determine the degree of change in the weather environment corresponding to the weather environment data of the historical wind and solar power generation data, and use the intermediate training data and the degree of change in the weather environment as target training data;

[0243] The prediction sub-model to be trained determines the training predicted wind and solar power generation power according to the target training data;

[0244] Adjust the parameters of the prediction sub-model according to the difference between the training predicted wind and solar power generation power and the actual wind and solar power generation power of the historical wind and solar power generation data, and update the weights corresponding to the hidden state information and the to-be-hidden state information;

[0245] Return to execute the step of removing the weighted processing result of the hidden state information from the historical wind and solar power generation data to obtain intermediate training data until the preset training conditions are met, and obtain the prediction sub-model corresponding to the category.

[0246] Each module in the above energy system control device considering wind and solar power generation power can be implemented in whole or in part by software, hardware and their combination. The above modules can be embedded in the processor in the computer device in hardware form or be independent of it, or can be stored in the memory in the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0247] In an exemplary embodiment, a computer device is provided. The computer device can be an energy storage terminal, and its internal structure diagram can be as Figure 7As shown in the figure. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface, the display unit, and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC), or other technologies. When the computer program is executed by the processor, it implements an energy system control method considering the power of wind and solar power generation. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0248] Those skilled in the art can understand that Figure 7 the structure shown in the figure is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0249] In one embodiment, a computer device is provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0250] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0251] In one embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0252] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data that have been authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0253] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., and are not limited thereto. The processors involved in the embodiments provided in this application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, artificial intelligence (AI) processors, etc., and are not limited thereto.

[0254] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as within the scope recorded in this application.

[0255] The above-described embodiments merely represent several implementation manners of this application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of this application. It should be noted that for those of ordinary skill in the art, without departing from the concept of this application, several modifications and improvements can still be made, and these all belong to the protection scope of this application. Therefore, the protection scope of this application shall be subject to the appended claims.

Claims

1. A control method for an energy system considering the power of wind and solar power generation, characterized in that, The method includes: Obtaining wind-solar power generation data; the wind-solar power generation data includes current wind-solar power generation power data and weather environment data; Extracting the time series features of the wind-solar power generation data and the correlation between the current wind-solar power generation power data and the weather environment data through a prediction model, and predicting the wind-solar power generation power according to the time series features and the correlation to obtain the predicted wind-solar power generation power; Obtaining the operating states of multiple energy units of the energy system; Based on the predicted wind-solar power generation power and the operating states, controlling the multiple energy units under the control instructions in the control instruction set to determine the operating resources consumed by the energy system; Taking minimizing the operating resources as the goal, determining the target control instructions for the multiple energy units from the control instruction set.

2. The method according to claim 1, characterized in that, The controlling the multiple energy units under the control instructions in the control instruction set based on the predicted wind-solar power generation power and the operating states to determine the operating resources consumed by the energy system includes: Determining a first operating state and a second operating state in the operating states; the first operating state is the operating state occurring at the next moment of the second operating state; Determining the working trend of the energy system according to the first operating state and the second operating state; Based on the predicted wind-solar power generation power and the working trend, controlling the multiple energy units under the control instructions in the control instruction set to determine the operating resources consumed by the energy system.

3. The method according to claim 2, wherein The taking minimizing the operating resources as the goal and determining the target control instructions for the multiple energy units from the control instruction set includes: Obtaining the operating parameters of the energy unit under the control instruction and comparing the operating parameters with a preset parameter threshold to obtain a comparison result; the comparison result is used to indicate the control priority of the energy unit; Based on the control priority, taking minimizing the operating resources as the goal, determining the target control instructions for the multiple energy units from the control instruction set; the target control instructions include turning on and off the energy unit.

4. The method according to claim 1, wherein The prediction model includes an encoder network and a decoder network; The extracting the time series features of the wind-solar power generation data and the correlation between the current wind-solar power generation power data and the weather environment data through a prediction model, and predicting the wind-solar power generation power according to the time series features and the correlation to obtain the predicted wind-solar power generation power includes: Inputting the wind-solar power generation data into the encoder network to extract an encoding result representing the time series features; Obtaining the weather forecast data of the weather environment data in a preset time period, inputting the encoding feature composed of the encoding result and the weather forecast data into the decoder network, determining the intermediate feature corresponding to the encoding feature by the intermediate layer of the decoder network, and obtaining a fusion feature by fusing the encoding feature and the intermediate feature; the fusion feature indicates the correlation between the current wind-solar power generation power data and the weather environment data; Inputting the fusion feature into the fully connected layer of the decoder network to output the predicted wind-solar power generation power.

5. The method according to claim 1, characterized in that, The prediction model is obtained through the following steps: Obtain historical wind and solar power generation data, and classify the historical wind and solar power generation data according to preset classification conditions to obtain historical wind and solar power generation data of multiple categories; For the historical wind and solar power generation data of each category, perform model training according to the historical wind and solar power generation data of the category to obtain a prediction sub-model corresponding to the historical wind and solar power generation data of the category; Integrate multiple prediction sub-models to obtain the prediction model; After obtaining the wind and solar power generation data, it further includes: Determine the wind and solar power generation data belonging to different categories, and predict the wind and solar power generation power corresponding to the wind and solar power generation data of each category according to the prediction sub-models corresponding to the wind and solar power generation data of different categories to obtain predicted wind and solar power generation power.

6. The method according to claim 5, wherein The performing model training according to the historical wind and solar power generation data of the category to obtain a prediction sub-model corresponding to the historical wind and solar power generation data of the category includes: Input the historical wind and solar power generation data into the prediction sub-model to be trained. The prediction sub-model to be trained determines the hidden state information corresponding to the time series characteristics in the historical wind and solar power generation data, and removes the weighted processing result of the hidden state information from the historical wind and solar power generation data to obtain intermediate training data; Determine the degree of change in the weather environment corresponding to the weather environment data of the historical wind and solar power generation data, and use the intermediate training data and the degree of change in the weather environment as target training data; The prediction sub-model to be trained determines the training predicted wind and solar power generation power according to the target training data; Adjust the parameters of the prediction sub-model according to the difference between the training predicted wind and solar power generation power and the actual wind and solar power generation power of the historical wind and solar power generation data, and update the hidden state information and the weights corresponding to the to-be-hidden state information; Return to execute the step of removing the weighted processing result of the hidden state information from the historical wind and solar power generation data to obtain intermediate training data until the preset training conditions are met, and obtain the prediction sub-model corresponding to the category.

7. An energy system control device considering the power of wind and solar power generation, characterized in that, The device includes: A wind and solar power generation data acquisition module for acquiring wind and solar power generation data; the wind and solar power generation data includes current wind and solar power generation power data and weather environment data; A wind and solar power generation power prediction module for extracting the time series characteristics of the wind and solar power generation data and the correlation between the current wind and solar power generation power data and the weather environment data through a prediction model, and predicting the wind and solar power generation power according to the time series characteristics and the correlation to obtain predicted wind and solar power generation power; An energy unit operation state acquisition module for acquiring the operation states of multiple energy units of the energy system; An operation resource determination module for controlling the multiple energy units under the control instructions in the control instruction set based on the predicted wind and solar power generation power and the operation states, and determining the operation resources consumed by the energy system; A target control instruction determination module for determining the target control instructions for the multiple energy units from the control instruction set with the goal of minimizing the operation resources.

8. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method described in any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method described in any one of claims 1 to 6.

Images