Hybrid energy storage optimization scheduling method, electronic equipment and computer readable storage medium

Abstract

The embodiment of the invention is suitable for the field of energy storage optimization scheduling, and provides a hybrid energy storage optimization scheduling method, electronic equipment and a computer readable storage medium, and the method comprises the steps: obtaining an electric power parameter of a hybrid energy storage system and a power grid operation parameter of a power grid; obtaining a first sequence composed of the generating capacity prediction data corresponding to the new energy power generation equipment at each future time point, and a second sequence composed of each load prediction data; based on the first sequence, the second sequence and the time-of-use electricity price of the power grid, establishing a target function which takes the charging and discharging power at each future time point as an unknown quantity and takes the charging and discharging cost as an optimization target; solving the target function by adopting a particle swarm algorithm to obtain target charging and discharging power at each future time point; and scheduling charging and discharging of the hybrid energy storage system based on the target charging and discharging power at each future time point. By the adoption of the method, a scientific optimization scheme can be provided for scheduling of the hybrid energy storage system, and the optimal economic benefit is achieved on the premise of safe operation.

Description

Technical Field

[0001] This application belongs to the field of energy storage optimization scheduling technology, and particularly relates to a hybrid energy storage optimization scheduling method, electronic device and computer-readable storage medium. Background Technology

[0002] With the development of smart grids and distributed energy storage technologies, distributed energy storage systems have become a potential solution to the grid regulation pressure brought about by the growth in electricity demand and the widening peak-valley difference. These systems offer bidirectional power output, rapid response, and flexible installation.

[0003] Currently, the regulation of hybrid energy storage systems that include multiple energy storage devices often adopts a fixed charge and discharge strategy. For example, based on the power system operating conditions (e.g., historical data such as grid power limits, load characteristics, energy storage parameters, and new energy power generation characteristics), combined with the functional positioning of the hybrid energy storage system (such as peak-valley arbitrage), a fixed charge and discharge power is manually set for a fixed period of time.

[0004] However, the aforementioned fixed charging and discharging strategy cannot dynamically adapt to the remaining energy storage capacity of the hybrid energy storage system (for example, it is difficult to maximize the utilization of energy storage potential when charging / discharging space is insufficient), and it cannot optimize energy allocation based on the load provided by new energy equipment. Consequently, the scheduling method of the hybrid energy storage system is not scientific. Summary of the Invention

[0005] This application provides a hybrid energy storage optimization scheduling method, electronic device, and computer-readable storage medium, which can solve the problem that the scheduling method of hybrid energy storage systems is not scientific.

[0006] In a first aspect, embodiments of this application provide a hybrid energy storage optimized scheduling method, the method comprising: Obtain the power parameters of the hybrid energy storage system and the grid operation parameters of the power grid; the power parameters are used to describe the physical characteristics of the hybrid energy storage system; the grid operation parameters include the maximum safe power during grid operation and the time-of-use electricity price. The operating constraints of the hybrid energy storage system are constructed based on power parameters and maximum safe power. A first sequence of predicted power generation data for new energy power generation equipment at each future time point is obtained, along with a second sequence of predicted load data. Based on the first sequence, the second sequence, and the grid time-of-use electricity price, an objective function is established with the charging and discharging power of the hybrid energy storage system at each future time point as the unknown quantity and the charging and discharging cost as the optimization objective. The objective function is solved using the particle swarm optimization algorithm to obtain the target charge and discharge power of the hybrid energy storage system at each future time point. During the process of solving the objective function, the charge and discharge power and the change in the amount of electricity of the hybrid energy storage system satisfy the operating constraints. At each future point in time, the charging and discharging of the hybrid energy storage system is scheduled based on the target charging and discharging power.

[0007] Secondly, embodiments of this application provide a hybrid energy storage optimized scheduling device, the device comprising: The first acquisition module is used to acquire the power parameters of the hybrid energy storage system and the grid operation parameters of the power grid; the power parameters are used to describe the physical characteristics of the hybrid energy storage system; the grid operation parameters include the maximum safe power of the power grid during operation and the time-of-use electricity price of the power grid. The construction module is used to construct the operating constraints of the hybrid energy storage system based on power parameters and maximum safe power. The second acquisition module is used to acquire a first sequence of power generation forecast data corresponding to each future time point of the new energy power generation equipment, and a second sequence of load forecast data. A module is established to create an objective function based on the first sequence, the second sequence, and the grid time-of-use electricity price, with the charging and discharging power of the hybrid energy storage system at each future time point as the unknown and the charging and discharging cost as the optimization objective. The solution module is used to solve the objective function using the particle swarm optimization algorithm to obtain the target charge and discharge power of the hybrid energy storage system at each future time point; during the process of solving the objective function, the charge and discharge power and the change in the amount of electricity of the hybrid energy storage system satisfy the operating constraints. The scheduling module is used to schedule the charging and discharging of the hybrid energy storage system based on the target charging and discharging power at various future time points.

[0008] Thirdly, embodiments of this application provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in the first aspect above.

[0009] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect above.

[0010] Fifthly, embodiments of this application provide a computer program product that, when run on an electronic device, causes the electronic device to execute the method described in the first aspect.

[0011] The beneficial effects of this application embodiment compared with the prior art are as follows: By obtaining the power parameters used to describe the physical characteristics of the hybrid energy storage system, and the maximum safe power during grid operation, operational constraints that conform to the operation of the hybrid energy storage system can be constructed. This ensures that the charging and discharging power and power changes of the hybrid energy storage system meet the operational constraints during the subsequent solution of the objective function. Subsequently, a first sequence of predicted power generation data corresponding to each future time point of the new energy power generation equipment and a second sequence of predicted load data can be obtained. Based on the first sequence, the second sequence, and the grid time-of-use pricing, an objective function is established with the charging and discharging power of the hybrid energy storage system at each future time point as the unknown quantity and the charging and discharging cost as the optimization objective. The objective function is then solved using a particle swarm optimization algorithm to obtain the target charging and discharging power of the hybrid energy storage system at each future time point. Based on this, the optimal charging and discharging power can be searched within the operational constraints using a particle swarm optimization algorithm, balancing safe operational constraints with the multi-dimensional objective of minimizing cost. Furthermore, by clearly defining low cost as the core objective through the objective function, and combining it with the grid time-of-use pricing to guide the charging and discharging control of the energy storage system, the high cost problem caused by the inability to dynamically adapt to price fluctuations in traditional fixed strategies can be solved. Finally, scheduling the charging and discharging of the hybrid energy storage system based on the target charging and discharging power at various future time points can provide a scientific optimization scheme for the scheduling of the hybrid energy storage system, achieving the best economic benefits under the premise of safe operation. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart illustrating the implementation of a hybrid energy storage optimization scheduling method according to an embodiment of this application; Figure 2 This is a schematic diagram illustrating one implementation method for generating target charging and discharging power in a hybrid energy storage optimization scheduling method provided in an embodiment of this application; Figure 3 This is a schematic diagram illustrating an implementation of iterative particles in a hybrid energy storage optimization scheduling method provided in an embodiment of this application; Figure 4 This is a schematic diagram illustrating one implementation method for determining correction coefficients in a hybrid energy storage optimization scheduling method provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of a hybrid energy storage optimized scheduling device provided in an embodiment of this application; Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0014] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0015] It should be noted that the information collection / feature extraction process involved in this application is carried out with the user's knowledge and permission, that is, the information collection / feature extraction process complies with the requirements of laws and regulations and does not constitute an act that harms the public interest.

[0016] In order to optimize the charging and discharging control strategy of hybrid energy storage systems and make the scheduling of hybrid energy storage systems more scientific, this application provides a hybrid energy storage optimization scheduling method. This method can be applied to electronic devices such as laptops, ultra-mobile personal computers (UMPCs), and netbooks. This application does not impose any restrictions on the specific type of electronic device.

[0017] Please see Figure 1 , Figure 1 The following is a flowchart illustrating the implementation of a hybrid energy storage optimization scheduling method provided in an embodiment of this application. The method includes the following steps: S101. Obtain the power parameters of the hybrid energy storage system and the grid operation parameters of the power grid.

[0018] In one embodiment, the aforementioned power parameters are used to describe the physical characteristics of the hybrid energy storage system. Among these, the grid operation parameters include the maximum safe power output during grid operation and the time-of-use electricity price.

[0019] In one embodiment, the hybrid energy storage system described above is an energy storage system that includes multiple different energy storage devices. This embodiment does not limit the type or number of energy storage devices included. For example, the hybrid energy storage system described above can be an energy storage system that includes lithium iron phosphate energy storage devices and vanadium redox flow storage devices.

[0020] The aforementioned power parameters reflect the operational capabilities and limitations of the hybrid energy storage system and are fundamental data for ensuring the safe and efficient operation of the system. These power parameters include, but are not limited to, the rated capacity, charging / discharging power, charging / discharging efficiency, and upper and lower limits of SOC (State of Charge) for each energy storage device; specific limitations are not imposed on these parameters.

[0021] For example, taking a hybrid energy storage system including lithium iron phosphate energy storage devices and vanadium redox flow storage devices as an example, the above-mentioned power parameters include, but are not limited to, the rated capacity of the lithium iron phosphate energy storage devices. Rated power Rated capacity of vanadium redox flow storage equipment Rated power The SOC operating range set for lithium iron phosphate energy storage devices ~ The State of Charge (SOC) operating space set for vanadium redox flow storage devices ~ and the total installed capacity of new energy power sources It should be noted that, based on the aforementioned power parameters, the charging efficiency of lithium iron phosphate energy storage devices can also be indirectly obtained through calculation and data processing. Discharge efficiency Charging efficiency of vanadium redox flow storage devices Discharge efficiency .

[0022] Specifically, electronic devices can use historical charging and discharging power and changes in SOC as a benchmark to calculate the actual charging and discharging power of each energy storage device.

[0023] For example, for lithium iron phosphate energy storage devices, electronic devices can use front-end metering sensors to measure the power consumption of historical charging. Historical discharge amount Correspondingly, based on the rated capacity and SOC change of the energy storage device, the actual charging capacity of the energy storage device is calculated to be: The actual discharge amount is For vanadium redox flow storage devices, electronic equipment can also measure the historical charging power consumption through front-end metering sensors. Historical discharge amount Correspondingly, based on the rated capacity and SOC change of the energy storage device, the actual charging capacity of the energy storage device is calculated as follows: The actual discharge amount is Furthermore, based on the above power parameters, the following can also be calculated: Charging efficiency of lithium iron phosphate energy storage devices ; Discharge efficiency of lithium iron phosphate energy storage devices ; Charging efficiency of vanadium redox flow storage devices ; Charging efficiency of lithium iron phosphate energy storage devices ; The SOC operating range and corresponding power capacity of lithium iron phosphate energy storage devices: ; ; The State of Charge (SOC) operating range and corresponding power capacity of vanadium redox flow storage devices: ; ; Among them, the rate of change of SOC during the charging process of lithium iron phosphate energy storage devices Rate of change of SOC during discharge process The planned charge and discharge power of the lithium iron phosphate energy storage device (the charge and discharge power that appears in the process of solving the objective function). The mathematical relationship; the rate of change of SOC during the charging process of a vanadium redox flow storage device. The planned charge and discharge power of the State of Charge (SOC) during the discharge process (charge and discharge power occurring during the solution of the objective function). The mathematical relationship is as follows: ; ; ; ; In one embodiment, the aforementioned power parameters can be obtained directly from the technical manuals of devices such as energy storage batteries and energy storage converters (PCS), such as rated capacity, maximum charge and discharge power, and efficiency. Alternatively, parameters such as charge and discharge efficiency and SOC upper and lower limits can be calibrated through actual charge and discharge tests. Or, parameters such as current SOC and instantaneous charge and discharge power can be collected in real time through the energy management system (EMS) or sensors of the energy storage system. Or, the aforementioned power parameters can be obtained through reading and writing databases, text input, or other means. There are no limitations on this method.

[0024] In one embodiment, the aforementioned power grid operation parameters may include not only the maximum safe power during power grid operation and the time-of-use electricity price, but also parameters such as the division of peak and valley periods of electricity load and reserve capacity requirements, which are not limited thereto.

[0025] The maximum safe power for grid operation refers to the maximum active power (unit: kW) allowed to pass through the grid during a specific time period. It is a core constraint for ensuring the stable operation of the grid. Exceeding this value may lead to line overload, voltage / frequency fluctuations, or even power outages. The aforementioned time-of-use pricing for the grid is a differentiated pricing system (unit: yuan / kWh) set by the grid based on the differences in load demand at different times. It is usually divided into peak hours (high load, high price), normal hours (medium load, medium price), and off-peak hours (low load, low price), with the aim of guiding users to use electricity during off-peak hours and optimizing the grid load curve.

[0026] In one embodiment, the aforementioned power grid operating parameters can be formulated by the local power dispatching agency based on factors such as power grid topology, peak load, and renewable energy access limits, and distributed through a dedicated dispatching interface or data platform. Furthermore, electronic devices can obtain the power grid operating parameters based on the dedicated dispatching interface or data platform. Alternatively, the aforementioned power grid operating parameters can be obtained through methods such as reading and writing databases or text input; there is no limitation on this method.

[0027] S102. Construct the operating constraints of the hybrid energy storage system based on power parameters and maximum safe power.

[0028] In one embodiment, the above-mentioned operational constraints are a series of limiting rules to ensure the safe, stable and efficient operation of the hybrid energy storage system during charging and discharging, and to coordinate with the power grid. These operational constraints form the basis for optimized scheduling.

[0029] In one embodiment, the aforementioned operational constraints include, but are not limited to, charging / discharging constraints, state of charge variation range constraints, and energy conservation constraints, etc., and are not limited thereto. Furthermore, these operational constraints can be set by the user or existing operational constraints can be used.

[0030] As an example, taking a hybrid energy storage system including lithium iron phosphate energy storage devices and vanadium redox flow storage devices as an example, its operating constraints can be as follows: Lithium iron phosphate (LFP) energy storage devices and vanadium redox flow storage devices are mutually exclusive in terms of charging and discharging; charging and discharging cannot occur simultaneously. The planned charging and discharging power of LFP and vanadium redox flow storage devices are also specified. , : ; The operating ranges of the State of Charge (SOC) for lithium iron phosphate energy storage devices and vanadium redox flow storage devices are as follows: ~ , ~ The initial SOC is its minimum SOC; the power supply constraint of new energy power generation equipment is that the power supply of new energy power generation equipment shall not exceed the installed capacity. Power grid constraints: The maximum power grid output shall not exceed the maximum safe power output. Power balance constraint, grid power = ; Planned charge and discharge balance constraints for lithium iron phosphate energy storage devices and vanadium redox flow storage devices; Local consumption constraints for new energy power generation devices, such as the minimum grid power when new energy power generation devices are not connected to the grid.

[0031] S103. Obtain a first sequence of power generation forecast data for each future time point of the new energy power generation equipment, and a second sequence of load forecast data.

[0032] In one embodiment, the aforementioned new energy power generation equipment refers to equipment that generates electricity using new energy sources such as wind power, solar power, hydropower, and biomass energy, such as wind turbines, solar photovoltaic panels, and hydroelectric generator sets. The aforementioned future time points can be any time point one day later, or any time point one week later, without limitation. The interval between any two adjacent future time points can be the same. For example, taking the next day as a future time point, the 24 hours of the next day can be divided into future time points with a granularity of 15 minutes (i.e., interval length). In this case, the number of future time points would be 96. The aforementioned power generation prediction data is a prediction of the electrical energy output that the new energy power generation equipment may generate at various future time points. Since new energy power generation is greatly affected by natural conditions (such as wind speed, light intensity, and water flow speed), and has strong uncertainty, this data can be obtained through prediction methods. The aforementioned first sequence is a data sequence composed of the power generation prediction data of the new energy power generation equipment at each future time point arranged in chronological order. For example, if we predict the solar power generation every 15 minutes over the next 24 hours, then the sequence consisting of 96 predicted power generation data points constitutes the first sequence. This predicted power generation data can be characterized by power generation output.

[0033] The aforementioned load forecast data represents predicted power load demand for the power system at various future points in time. Power load demand is influenced by various factors, such as user electricity consumption habits, seasonal variations, weather conditions, and industrial production activities. This load forecast data can be characterized by load power. The second sequence is a data sequence composed of various load forecast data arranged in chronological order. Similar to the first sequence, it is also organized based on future points in time. For example, the power load forecast data for every 15 minutes within the next 24 hours constitutes the second sequence.

[0034] In one embodiment, power generation forecast data can be obtained through model prediction using meteorological data at future points in time. For example, electronic devices can collect meteorological data using devices such as meteorological satellites and radars, and then use numerical weather prediction models to predict meteorological conditions at each future point in time. Subsequently, based on the characteristics of the new energy power generation equipment (such as the power curve of a wind turbine, the photoelectric conversion efficiency curve of a solar photovoltaic panel, etc.), combined with the predicted meteorological data, the predicted power generation value of the new energy power generation equipment at each future point in time can be calculated. For example, for wind power generation, based on the predicted wind speed and the power curve of the wind turbine, the power generation at different wind speeds can be calculated. Alternatively, electronic devices can collect historical power generation data of the new energy power generation equipment, along with corresponding meteorological data, time information, etc., to establish a statistical model between power generation and related influencing factors (meteorological data and time information). Then, using this statistical model, combined with future meteorological forecast data and time information, the power generation at each future point in time can be predicted. For example, by analyzing meteorological data such as wind speed and sunshine duration over the same period in the past few years, and the corresponding power generation data, a regression model can be established, and then the future meteorological forecast data can be input into the model to obtain the predicted power generation value (i.e., power generation forecast data). In this embodiment, the method of obtaining power generation prediction data is not limited.

[0035] In one embodiment, the method for obtaining the load forecast data can be similar to the method for obtaining the power generation forecast data. For example, the load forecast data can also be obtained based on a pre-trained load forecast model, which will not be described in detail here.

[0036] S104. Based on the first sequence, the second sequence, and the grid time-of-use electricity price, establish an objective function with the charging and discharging power of the hybrid energy storage system at each future time point as the unknown quantity and the charging and discharging cost as the optimization objective.

[0037] In one embodiment, the objective function is a mathematical function constructed using a first sequence of new energy power generation forecast data, a second sequence of load forecast data, and the time-of-use electricity price of the power grid as basic information. In the objective function, the charging and discharging power of the hybrid energy storage system at each future time point can be considered as the unknown variable to be solved, and the charging and discharging cost during the operation of the hybrid energy storage system can be considered as the target to be optimized, thus determining the sequence of target charging and discharging power with the optimal charging and discharging cost.

[0038] The objective function mentioned above can be a nonlinear fitting function or an integer programming function, and there are no restrictions on this.

[0039] In one embodiment, the aforementioned charging and discharging costs may include one or more of the following: grid interaction costs, flip-flop costs, and device lifespan costs. The grid interaction costs refer to the costs or benefits incurred when the hybrid energy storage system interacts with the grid (charging or discharging). For example, charging costs are the electricity fees paid when the hybrid energy storage system charges from the grid. This cost is related to the grid's time-of-use pricing, typically lower during off-peak hours and higher during peak hours. For instance, at night when grid load is low, the grid's time-of-use pricing may be lower, resulting in relatively lower charging costs for the hybrid energy storage system; conversely, during peak daytime electricity demand, the grid's time-of-use pricing is higher, leading to increased charging costs.

[0040] Discharge revenue: The revenue generated from electricity sales when the hybrid energy storage system discharges into the grid. Discharge revenue is also affected by the grid's time-of-use pricing; discharging during peak electricity price periods yields higher revenue. For example, releasing stored energy into the grid during peak daytime electricity demand can generate higher revenue than during off-peak periods.

[0041] The aforementioned switching costs refer to the costs incurred when the various energy storage devices in a hybrid energy storage system switch between charging and discharging states (i.e., switchover). For example, control costs: switching charging and discharging states requires adjustment and operation of the corresponding control system, which involves the operation and maintenance costs of the control equipment. For instance, the energy consumption of the control circuit, upgrades and maintenance of the control software all require certain expenses. Time costs: The state switching process takes a certain amount of time. During this time, the hybrid energy storage system may not be able to perform normal charging and discharging operations, thus affecting its economic benefits. For example, during the switching process, the hybrid energy storage system may need to stop charging and discharging for a period of time, causing it to miss the discharge opportunity during peak electricity price periods, resulting in lost revenue.

[0042] The aforementioned equipment lifespan cost refers to the costs incurred by a hybrid energy storage system due to lifespan degradation during use. For example, battery health status cost: the cost resulting from the difference between the battery health status of each energy storage device at various future points in time and a standard reference health status. Alternatively, maintenance cost: the cost of regularly maintaining and servicing the energy storage devices to ensure their normal operation and extend their lifespan.

[0043] As an example, when the charging and discharging cost only includes grid interaction costs, it can be calculated using the following formula: ; in, Let T represent the power grid interaction cost, T be the total number of future time points, and t be the t-th future time point. This is the predicted power generation data for the t-th future time point. For the load forecast data at the t-th future time point, Let be the charging and discharging power of the first type of energy storage device at the future time point t. Let be the charging and discharging power of the second type of energy storage device at the t-th future time point, and step be the preset time period corresponding to two adjacent future time points. Let be the time-of-use electricity price for the power grid at the t-th future time point.

[0044] As another example, when the charging and discharging cost includes grid interaction costs and switching costs, it can be calculated using the following formula: ; in, This represents the number of charge-discharge reversals of the first type of energy storage device at all future time points. This represents the initial switching cost corresponding to each charge / discharge cycle of the first type of energy storage device. This represents the number of charge-discharge reversals for the second type of energy storage device at all future time points. This represents the initial switching cost corresponding to each charge / discharge cycle of the second type of energy storage device.

[0045] It should be noted that in the above calculation formula... and The unknown quantity in the objective function is the charging and discharging power that needs to be solved. However, when calculating the charging and discharging cost using the above two formulas, the losses of each energy storage device in the hybrid energy storage system during the charging and discharging process are ignored. Furthermore, for the part of the formula that calculates the switching cost in the second formula, it is simply calculated by multiplying the number of charging and discharging switching by the initial switching cost. This calculation method is too simplistic and does not take into account that the actual losses caused to the equipment may be different under different charging and discharging state switching situations (such as the grid time-of-use price during switching, switching under different load conditions). For example, the degree of equipment loss may be different when performing charging and discharging switching under high load conditions and low load conditions, but this calculation method cannot reflect this difference, resulting in an inaccurate calculation of the switching cost, which in turn affects the accuracy of the overall charging and discharging cost calculation and the scientific nature of the scheduling strategy. Based on this, in order to accurately evaluate the charging and discharging cost, taking a hybrid system including two types of energy storage devices as an example, the above objective function can be as follows: ; in, For charging and discharging costs, Indicates the cost of grid interaction. This indicates the switching cost of energy storage devices during charge / discharge cycles. Let T be the equipment lifespan cost of the energy storage device, T be the total number of future time points, and t be the t-th future time point. This is the predicted power generation data for the t-th future time point. For the load forecast data at the t-th future time point, Let be the charging and discharging power of the first type of energy storage device at the future time point t. Let be the charging and discharging power of the second type of energy storage device at the t-th future time point, and step be the preset time period corresponding to two adjacent future time points. Let t be the time-of-use electricity price for the power grid at the t-th future time point. Let be the number of charge-discharge reversals of the first type of energy storage device at a future time point t. Let $\frac{ ... Let be the number of charge-discharge reversals for the second type of energy storage device at the future time point t. Let $\frac{ ... These represent preset coefficients. This represents the health status of the first type of energy storage device at a future time point t. This is the reference health status for the first type of energy storage device. The health status of the second type of energy storage device at a future time point t. This is a reference health status for the second type of energy storage device.

[0046] In one embodiment, in a power system, generation power, load power, and the charging and discharging power of energy storage systems need to be balanced to ensure the stable operation of the power grid. Specifically, in the above calculation formula, for each future time point t, the following power balance relationship exists: ; in, This refers to the power (charging power or discharging power) that the hybrid energy storage system interacts with the grid.

[0047] It should be noted that the above formula is used to calculate the grid interaction cost. The value may be negative. When negative, it can be assumed that the hybrid energy storage system can supply power to the grid, thus generating revenue for the system. In this case, the grid interaction cost corresponding to that future time point will be negative. When this negative value is added to the grid interaction costs corresponding to other future time points, the overall charging and discharging cost can be reduced. That is, the charging and discharging cost may be negative. This characteristic allows the solution to be performed with minimizing the charging and discharging cost as the optimization objective.

[0048] Understandably, this involves the power interacting with the power grid. Multiplying by the time-of-use electricity price at future time t allows us to calculate the cost or benefit of interacting with the grid at that time. Furthermore, This involves converting the time step to hours to match the electricity price unit (yuan / kWh). For example, if the time step (the preset time period corresponding to two adjacent time points) is 15 minutes, it corresponds to 0.25 hours.

[0049] It should be noted that the preset time interval corresponding to the two adjacent time points mentioned above can be a time interval shorter than the preset value. For example, 15 minutes. By using a smaller time step to acquire data, the number of future time points can be increased to find a better charging and discharging strategy. Based on this, according to the above calculation formula... By maintaining power balance, considering time-of-use pricing mechanisms, and minimizing interaction costs, the economically optimal interaction between hybrid energy storage systems and the power grid can be achieved.

[0050] In one embodiment, the above and It can be dynamically set according to actual factors. For example, the above... and The future time-of-use (TOU) electricity price can be determined based on the corresponding TOU price and the charging / discharging state after the reversal. The TOU price for the future time point is determined according to the grid's TOU policy. For example, the electricity price during the day is divided into three periods: peak, flat, and off-peak. The peak period price is set as the high price, and its corresponding price range is determined. Simultaneously, during the charging / discharging scheduling process, the state (discharging or charging) of the energy storage device after each charging / discharging reversal operation in the hybrid energy storage system is assessed. If the TOU price is higher than a threshold and the device is in a discharging state after the reversal, then... and Reduce the preset percentage to lower the initial switching cost at that future time point. If the grid time-of-use price is higher than the threshold and the device is in a charging state after the switching, then... and Increase the preset percentage to increase the initial switching cost corresponding to that future time point. If the grid time-of-use price is lower than or equal to the threshold, it can be maintained. and constant.

[0051] Understandably, in the discharge state, the hybrid energy storage system can supply power to the grid to offset peak electricity demand, thus increasing revenue during this switching operation. Therefore, the initial switching cost can be reduced. Conversely, in the charging state, the hybrid energy storage system may need to store electricity output from the grid. Consequently, it needs to pay costs to the grid, thus increasing costs during this switching operation. Based on this, the initial switching cost can be increased.

[0052] In another embodiment, the electronic device can also adjust the initial switching cost based on the predicted charging and discharging power at various future time points. For example, high-power charging and discharging operations may cause greater stress on the electrical and mechanical components of the energy storage device, increasing its wear and maintenance costs. Therefore, when the charging and discharging power at the corresponding future time point exceeds a certain threshold, the initial switching cost can be increased by a preset percentage. Otherwise, the initial switching cost remains unchanged. In this embodiment, the method of dynamically setting the initial switching cost at each future time point is not limited.

[0053] Based on the above explanation, the calculation formula is adopted. Calculating the cost of switching over time can not only capture cost changes at different times, but also reflect the differences in characteristics of different energy storage devices, making cost calculations more realistic.

[0054] In one embodiment, the above and This can be achieved through machine learning methods. For example, a large amount of historical operating data (including weather, charging / discharging current, voltage, temperature, and charge / discharge cycle counts) and corresponding State of Health (SOH) values ​​(partial samples obtained through experimental testing) can be used to train a machine learning model. Examples include neural networks (Long Short-Term Memory networks, etc.) and support vector machines. The real-time collected operating data is then input into the trained model to predict the SOH at various future time points. For instance, using a Long Short-Term Memory network to train on historical lithium battery data allows the model to learn the complex mapping relationship between operating parameters and SOH, thereby accurately predicting the SOH at different time points.

[0055] It should be noted that the calculation formula is used. Calculating equipment lifecycle cost can be achieved by using the squared term in the formula to accurately measure the deviation in equipment health status, and by adding it over time to comprehensively reflect the lifecycle loss of each energy storage device throughout the entire cycle, making the obtained equipment lifecycle cost more accurate.

[0056] In this embodiment, by calculating the grid interaction costs arising from the interaction between the hybrid energy storage system and the grid, the switching costs due to charge / discharge state transitions, and the equipment lifespan costs resulting from long-term equipment use, a more comprehensive reflection of actual operating costs can be achieved compared to calculation methods that only consider a single cost type. This avoids decision-making biases caused by incomplete cost considerations. Furthermore, the calculation process fully considers the impact of time factors on costs. For example, the time-of-use electricity price in grid interaction costs changes over time, the initial switching cost in switching costs can be dynamically adjusted according to time, and equipment lifespan costs are related to the health status of the equipment at different points in time. Based on the dynamic impact of time factors on costs, the calculation of charge / discharge costs can reflect changes in system operating conditions in real time, improving the accuracy and timeliness of charge / discharge cost calculations.

[0057] In one embodiment, the first type of energy storage device can be a lithium iron phosphate energy storage device, and the second type of energy storage device can be a vanadium redox flow storage device. In this case, the preset coefficient corresponding to the lithium iron phosphate energy storage device is greater than the preset coefficient corresponding to the vanadium redox flow storage device, and the initial switching cost corresponding to the lithium iron phosphate energy storage device is greater than the initial switching cost corresponding to the vanadium redox flow storage device.

[0058] It should be noted that lithium iron phosphate (LFP) energy storage devices have a relatively limited cycle life, and their capacity typically shows significant degradation after a certain number of charge-discharge cycles. Vanadium redox flow storage (VRF) devices, on the other hand, have a longer cycle life and can withstand more charge-discharge cycles without significant performance degradation. Because of the relatively short lifespan of LFP devices, changes in their health status have a more critical impact on the overall operating cost of hybrid energy storage systems. Therefore, a larger preset coefficient is needed to highlight the importance of lifespan degradation in the calculation of device lifespan costs. For example, under the same number of charge-discharge cycles, the capacity degradation of LFP devices may reach 20%, while that of VRF devices may only be around 5%. This makes the lifespan degradation of LFP devices have a more significant impact on cost, requiring a larger preset coefficient to reflect this.

[0059] Furthermore, lithium iron phosphate (LFP) energy storage devices have a faster charge and discharge response speed, enabling them to complete high-power charge and discharge operations in a short time. In contrast, vanadium redox flow storage (VRF) devices have a relatively slower charge and discharge response speed, requiring a longer time to adjust power. Due to the faster response speed of LFP devices, the internal chemical reactions and electrode material changes are more drastic during charge / discharge cycles, leading to greater equipment losses and increased switching costs. For example, during rapid charge and discharge, lithium plating may occur on the electrode surface of LFP devices, accelerating electrode material aging. VRF devices, due to their slower response speed, experience less loss, thus resulting in a higher initial switching cost for LFP devices.

[0060] In summary, based on the differences between lithium iron phosphate (LFP) energy storage devices and vanadium redox flow storage devices in terms of cycle life and charge / discharge response speed, setting a higher preset coefficient for LFP energy storage devices and a higher initial switching cost for vanadium redox flow storage devices can more accurately reflect the actual costs and losses of the two types of energy storage devices during charge / discharge switching.

[0061] S105. The objective function is solved using the particle swarm optimization algorithm to obtain the target charging and discharging power of the hybrid energy storage system at each future time point.

[0062] In one embodiment, during the solution of the objective function, the charging and discharging power and energy changes of the hybrid energy storage system satisfy the operational constraints. The Particle Swarm Optimization (PSO) algorithm is a swarm intelligence-based optimization algorithm that simulates the collective behavior of organisms such as flocks of birds foraging. In the algorithm, each particle represents a solution (i.e., a charging and discharging power corresponding to a given future time point), and the optimal solution corresponds to the target charging and discharging power at each future time point. Particles move in the solution space, updating their positions by tracking individual and swarm optimal solutions, ultimately finding the global optimal solution.

[0063] In one embodiment, the process of solving the objective function can be referred to Figure 2 The steps shown are as follows: Initialize the particle swarm: Particle Count: The size of the particle swarm (i.e., the number of particles) is determined based on the complexity of the problem and the limitations of computational resources. A larger number of particles allows for a wider search range, but also increases computational complexity. Particle Dimension: The particle dimension can be determined based on the number of future time points. For example, to solve for the target charge / discharge power of two energy storage devices at 96 points (i.e., the charge / discharge status of the energy storage system every 15 minutes), the particle dimension for each energy storage device would be 96 dimensions, resulting in a total particle dimension of 192 dimensions. Particle Position: The position of each particle corresponds to a set of decision variables, i.e., the charge / discharge power of the hybrid energy storage system at each future time point. The particle position is randomly initialized in the solution space. Particle Velocity: The velocity of each particle determines its direction and step size in the solution space. The velocity is also randomly initialized, typically within a certain range.

[0064] Initialize particle constraint processing: The randomly initialized particles are corrected based on preset operating constraints to ensure that the initialized particles conform to the physical constraints of the hybrid energy storage system, thereby accelerating the convergence speed of the particle swarm algorithm.

[0065] Assess particle fitness: Calculate the objective function value: Substitute the position of each particle (i.e., a set of charging and discharging powers) into the objective function to calculate the corresponding charging and discharging cost. Fitness evaluation: Fitness is an indicator of a particle's quality. During the solution process, fitness can be directly expressed as the reciprocal or a negative number of the objective function value (because the goal is to minimize cost). The smaller the objective function value, the higher the fitness.

[0066] Update individual optimality and group optimality: Individual optimality: Each particle records its own optimal position found during the search process (i.e., the charging / discharging power corresponding to the minimum objective function value). Swarm optimality: Records the optimal positions found by all particles in the entire particle swarm.

[0067] Iteration termination condition determination: Output: When the termination condition is met, the charging and discharging power corresponding to the optimal position of the output group is the target charging and discharging power of the hybrid energy storage system at each future time point. These power values ​​satisfy the operating constraints and minimize the charging and discharging costs.

[0068] Iterative phase constraint handling: When the termination condition is not met, after updating the individual and group optima, it is necessary to check whether the positions corresponding to the individual and group optima satisfy the operational constraints. If not, the positions need to be adjusted, for example, adjusting the charging and discharging power exceeding the power limit to within the allowable range. The adjustment method can be as follows: for each position corresponding to a particle, if the charging and discharging power corresponding to the position is greater than the maximum critical value, then the maximum critical value is determined as the charging and discharging power corresponding to that particle. If the charging and discharging power corresponding to the position is less than the minimum critical value, then the minimum critical value is determined as the charging and discharging power corresponding to that particle.

[0069] For example, if we want to solve for the charging and discharging power of a 96-point hybrid energy storage system, the upper and lower limits in particle theory should be the maximum charging and discharging power of the corresponding energy storage device. However, the upper and lower limits of the charging and discharging power are usually different for different energy storage devices.

[0070] The above-mentioned particle correction process needs to take into account constraints such as the power grid's safety power, which will not be explained in detail here.

[0071] Update the particle's velocity and position: The velocity update involves updating the velocity of each particle based on its optimal individual and global positions that satisfy the constraints. The velocity update formula is typically: ; Where i represents the i-th particle, and t represents the t-th time. This represents the velocity of the i-th particle at time t+1. This represents the historical best position of the i-th particle. The global optimal position of the entire particle swarm, r1 and r2 are random numbers between 0 and 1 to increase the randomness of the search, C1 is the first learning factor for the particle to perform local search on its own historical optimal position, C2 is the second learning factor for the particle to perform global search on the global optimal position, and C3 is the third learning factor used to balance the local search and global search of the particle.

[0072] Position update: Update the position of each particle based on the updated velocity; Increment the iteration count by 1 and repeat the steps described above, from evaluating particle fitness to updating particle velocity and position, until the iteration termination condition is met. The termination condition includes, but is not limited to, the maximum number of iterations or the objective function value remaining unchanged for a predetermined number of consecutive iterations.

[0073] Based on the above description, in the particle swarm initialization phase and the iterative update phase for optimizing particles by solving the objective function, if a particle violates the operating constraints, a preset correction rule is used to correct the particle; the corrected particle satisfies the operating constraints; wherein, the particle represents a charging and discharging power at various future time points. Then, the charging and discharging power corresponding to the particle after the iteration phase ends is determined as the target charging and discharging power. The above correction rule can be set according to actual needs and is not limited thereto. For example, the method of correcting particles using the correction rule can refer to the above constraint processing content, which will not be described in detail here. Furthermore, the above iterative update phase may include the above... Figure 2 The process of evaluating particle fitness to update particle velocity and position.

[0074] Understandably, the pre-defined correction rules essentially transform random solutions at the algorithmic level into feasible solutions at the physical level. Through targeted corrections (e.g., truncation, reverse constraints, efficiency corrections, and collaborative complementarity), it can be ensured that the charging and discharging power of the particles is both within the algorithm's search boundary and conforms to the hardware characteristics, energy laws, and collaborative requirements of the energy storage system. The corrected particles can not only be further optimized by the particle swarm optimization algorithm but also directly correspond to practically executable scheduling strategies, laying a feasible foundation for subsequently determining the target charging and discharging power.

[0075] It should be noted that the operating constraints of a hybrid energy storage system are essentially a mapping of physical laws (such as SOC cannot be negative or exceed 100%, and charging / discharging power cannot exceed the rated range of the equipment). If the operating constraints are not met, even if the particle position is within the bounds set by the algorithm (e.g., [-500, 500]), scheduling strategies may still fail to execute.

[0076] For example, when the SOC of a lithium iron phosphate energy storage device is 0 (fully discharged), physical constraints require that its discharge power must be 0 (discharge cannot continue). If the particle positions are not corrected, they may be randomly initialized to -300 (representing a discharge of 300kW). Although this result conforms to the bounds in the algorithm, the device cannot execute in reality (no power to discharge). A scheduling strategy that meets the operating constraints will correct the discharge power to 0 at this time, ensuring that the scheduling strategy can be executed.

[0077] Furthermore, ensuring that the charging and discharging power and charge changes satisfy the operational constraints during particle swarm initialization and iteration allows the particles to always search within the physically feasible solution space, accelerating convergence to the true optimal solution. This, in turn, avoids wasting computation time in invalid regions during the solution process and ensures the validity of the optimization objective, preventing "false optimal solutions."

[0078] S106. At each future time point, schedule the charging and discharging of the hybrid energy storage system based on the target charging and discharging power.

[0079] In one embodiment, after obtaining the target charge / discharge power at each future time point, the target charge / discharge power at that future time point can be considered as an optimized scheduling strategy. Electronic devices can control each energy storage device in the energy storage system, operating based on the target charge / discharge power corresponding to each future time point, thereby enabling the energy storage system to achieve low charge / discharge costs and high returns.

[0080] In this embodiment, by acquiring the electrical parameters describing the physical characteristics of the hybrid energy storage system and the maximum safe power during grid operation, operational constraints that conform to the operation of the hybrid energy storage system can be constructed. This ensures that the charging and discharging power and power changes of the hybrid energy storage system satisfy the operational constraints during the subsequent solution of the objective function. Next, a first sequence of predicted power generation data for each future time point of the new energy power generation equipment and a second sequence of predicted load data can be obtained. Based on the first and second sequences and the grid time-of-use pricing, an objective function is established with the charging and discharging power of the hybrid energy storage system at each future time point as the unknown and the charging and discharging cost as the optimization objective. The objective function is then solved using a particle swarm optimization algorithm to obtain the target charging and discharging power of the hybrid energy storage system at each future time point. Based on this, the optimal charging and discharging power can be searched within the operational constraints using the particle swarm optimization algorithm, balancing safe operational constraints with the multi-dimensional objective of minimizing cost. Furthermore, by clearly defining low cost as the core objective through the objective function, and combining it with grid time-of-use pricing to guide the charging and discharging control of the energy storage system, the high cost problem caused by the inability to dynamically adapt to price fluctuations in traditional fixed strategies can be solved. Finally, scheduling the charging and discharging of the hybrid energy storage system based on the target charging and discharging power at various future time points can provide a scientific optimization scheme for the scheduling of the hybrid energy storage system, achieving the best economic benefits under the premise of safe operation.

[0081] In another embodiment, the iterative update phase described above includes the above-described... Figure 2 The process involves evaluating particle fitness and updating particle velocity and position. This process requires updates based on learning factors C1, C2, and C3. Currently, in related technologies, C1, C2, and C3 are typically pre-set fixed values. However, in the optimization problem of hybrid energy storage systems, the solution space of the objective function (the feasible region of charge and discharge power) may exhibit complex characteristics (e.g., non-convex, multi-peak) due to the passage of time and dynamic changes in constraints (such as SOC upper and lower limits, power upper and lower limits). In this case, fixed learning factors lack adaptability to the characteristics of the solution space.

[0082] Specifically, when the solution space is large (e.g., in the early stages of iteration when a wide range of feasible solutions needs to be explored), fixing a small C2 will restrict the particle's movement towards the global optimum, leading to insufficient exploration and the omission of better solutions. When the solution space is narrow (e.g., in the later stages of iteration when fine-tuning of charging and discharging power is needed to meet the smooth changes in SOC), fixing a large C1 will cause the particle to oscillate within a local range, failing to converge stably to the optimal solution. Therefore, in order to obtain a stably convergent optimal solution, the electronic device can, as shown in the example... Figure 3 The steps S301-S303 shown dynamically adjust the learning factor to iteratively update the particles. Details are as follows: S301. For any iteration in the iteration phase, obtain the current iteration number and the standard deviation of the particle position corresponding to the current iteration number.

[0083] In one embodiment, the aforementioned iterative phase can be considered as the phase other than the initialization phase and the iteration end phase. During the iterative phase, the particle gradually moves closer to the optimal solution. The current iteration number refers to the k-th iteration (k is a positive integer, counting from 1) in which the algorithm optimizes and updates the particle starting from the initialization phase. For example, if the algorithm is set to a total of 100 iterations, the current iteration number is 30 when the 30th update is executed.

[0084] In one embodiment, the electronic device can record data in real time based on a built-in counter. The counter automatically increments by 1 after each iteration. The position standard deviation of the particles corresponding to the current iteration number refers to a quantitative indicator of the dispersion (dispersion) of the positions of all particles in the solution space at the current iteration number. Since the "position" of each particle characterizes the charging and discharging power of the hybrid energy storage system at various future time points (e.g., a particle corresponds to a 24-hour charging and discharging power sequence, which can be considered a multi-dimensional vector), the position standard deviation can be calculated as follows: Single-dimensional standard deviation calculation method: For the charging and discharging power at each future time point (dimension), calculate the standard deviation of the power value of all particles in each dimension (reflecting the dispersion of power at that time point).

[0085] Alternatively, the global standard deviation calculation method treats the positions of all iterated particles as multi-dimensional vectors and calculates the Euclidean distance between vectors or the mean of the standard deviations of each dimension to comprehensively reflect the dispersion of the overall particle swarm.

[0086] As an example, taking single-dimensional standard deviation, the electronic device can obtain all positions of particles at the current iteration number (assuming there are N particles, each particle's position is a T-dimensional vector, where T is the number of future time points, such as T=96 in a 24-hour schedule). Then, power values ​​are extracted by dimension: for each future time point t (t=1,2,...,T), the charging and discharging power values ​​of all N particles in that dimension are extracted, resulting in a power value sequence: x{1,t}, x{2,t},...,x{N,t}, where x{i,t} is the charging and discharging power of the i-th particle at time t. Then, for each future time point t, the standard deviation corresponding to the power value sequence is calculated based on the standard deviation formula. Finally, the average of the standard deviations corresponding to each future time point is taken to obtain the aforementioned positional standard deviation.

[0087] It should be noted that the current iteration number reflects the search stage of the algorithm (e.g., the initial stage requires exploring a wide-area solution space, while the later stage requires fine-grained convergence), while the position standard deviation reflects the dispersion of the particle swarm (a large standard deviation indicates dispersed particles and strong exploration ability; a small standard deviation indicates clustered particles and a clear convergence trend). Based on this, by combining both and dynamically adjusting the learning factor, the algorithm can maintain particle dispersion in the early stages of iteration to expand the search range, and promote particle clustering in the later stages of iteration to improve convergence accuracy, thus better meeting the needs of optimizing the charging and discharging power of hybrid energy storage systems.

[0088] S302. Based on the position standard deviation and the current iteration number, determine the target value of the learning factor corresponding to the current iteration number.

[0089] In one embodiment, the electronic device may have a pre-set mapping relationship between position standard deviation, number of iterations and learning factor. Based on the above-mentioned pre-set mapping relationship, the target value of the learning factor corresponding to the current number of iterations can be determined.

[0090] In another embodiment, the electronic device can be based on, for example... Figure 4 Steps S401-S404, as shown, determine the target value of the learning factor. Details are as follows: S401. Calculate the ratio of the current iteration number to the preset total iteration number.

[0091] S402. Based on the difference and ratio between the maximum and minimum critical values ​​within the preset range corresponding to the learning factor, determine the initial value of the learning factor corresponding to the current iteration number.

[0092] In one embodiment, the aforementioned preset total number of iterations is a pre-set maximum number of iterations, which can be one of the conditions for terminating the algorithm. The aforementioned ratio can be used to quantify the relative progress of the current iteration in the entire optimization process. For example, a ratio of 0.25 indicates 25% progress, and 0.8 indicates 80% progress, providing a standardized timescale for the dynamic adjustment of the learning factor.

[0093] The aforementioned learning factors may include the first learning factor C1, the second learning factor C2, and the third learning factor C3 described above, without limitation. The preset range is a pre-defined interval for the learning factors, representing the upper and lower limits of the learning factor fluctuations allowed by the algorithm. For example, a preset range of [1.5, 2.5] indicates that the learning factor value must be between 1.5 and 2.5. Different learning factors may correspond to different preset ranges. It should be noted that constraining the dynamic adjustment range of the learning factors can prevent abnormal particle search behavior (such as excessive divergence or premature convergence) caused by learning factors that are too large or too small.

[0094] In one embodiment, the maximum threshold and the minimum threshold are the upper and lower limits of the preset range of the learning factor, respectively, to reflect the adjustable range of the learning factor.

[0095] As an example, an electronic device can determine the initial value of the learning factor corresponding to the current iteration number based on a preset mapping relationship between the difference, ratio, and the initial value of the learning factor.

[0096] In another embodiment, taking the learning factors as divided into a first learning factor, a second learning factor, and a third learning factor as an example, the initial value corresponding to each learning factor can be calculated according to the formula shown below. Details are as follows: The initial value of the first learning factor is calculated based on a preset formula; the formula for calculating the first learning factor is as follows: ; Where C1 is the initial value of the first learning factor. The maximum critical value corresponding to the first learning factor. d is the minimum critical value corresponding to the first learning factor, and d is the ratio.

[0097] It is understandable that, based on the above calculation formula, the introduction of hyperbolic tangent (tanh), sine (sin), and polynomial terms into the formula can cause the first learning factor C1 to exhibit dynamic nonlinear changes during the iteration process, thereby precisely controlling the particle's search behavior.

[0098] It should be noted that in the scenario of optimizing the charging and discharging power of hybrid energy storage systems, the particle swarm optimization algorithm has significantly different requirements for the "exploration-development" stages. Typically, in the initial stage, it needs to quickly cover the solution space (to avoid missing optimal solutions); in the middle stage, it needs to balance the search direction (to avoid getting trapped in local conditions); and in the later stage, it needs precise convergence (to improve the accuracy of the solution). In the above calculation formula, two composite functions can be used to specifically meet the requirements of each stage. Specifically, with... As the base value gradually increases, the value of C1 is relatively low in the early to mid-stages of the iteration, making the particles more inclined to explore extensively in the solution space and avoiding premature convergence.

[0099] It can undertake the task of "balancing exploration and convergence in the early and middle stages". This is understandable. The growth is faster when d is in the range [0, 0.3] (early stage of iteration). For The growth of d is also relatively rapid when it is in the range [0, 0.5] (mid-iteration stage). The product of the two factors can gradually increase the contribution to C1 in the early and mid-stages of the iteration. That is, C1 gradually increases, thereby balancing the particle's exploration and convergence capabilities. At this time, in the early and mid-stages of the iteration, the initial value corresponding to C1 is relatively small, and the particle relies more on the global optimum (second learning factor) and inertia weight (third learning factor) to explore extensively in the solution space, avoiding premature convergence to a local optimum.

[0100] For example, when d=0.2, tanh(0.6) is approximately 0.537, sin(0.314) is approximately 0.309, and the product is approximately 0.166. At this time, C1 is relatively small, and the particle is more inclined to explore. When d=0.5, tanh(1.5) is approximately 0.905, sin(0.785) is approximately 0.707, and the product is approximately 0.640. At this time, C1 is relatively large, and the particle focuses on both its own optimal (pbest) and the global optimal (gbest), and is more inclined to balance exploration and convergence.

[0101] And, for 0.1d 3 It can undertake the task of "late-stage stable convergence". Specifically, throughout the entire iteration process, as d increases, d... 3 The growth rate is gradually accelerating, 0.1d 3 The contribution of the term to C1 gradually increases. In the later stages of the iteration, when d is in [0.7, 1], the growth rate of the previous term gradually slows down, while the term can still provide a stable contribution, avoiding a sudden increase in C1 that could cause a sudden change in particle velocity and ensuring the stability of the convergence process.

[0102] For example, when d=0.8, 0.1d 3 = 0.0512; when d=0.9, 0.1d 3 = 0.0729, the contribution of this term gradually increases, causing C1 to continue to increase in the later stage, and thus the particles are more inclined to move towards their own optimal (pbest) and global optimal (gbest), which strengthens the local search and accelerates the convergence to the optimal solution.

[0103] Based on this, the initial value C1 of the first learning factor is calculated using the above formula, which can simulate the natural search law of the particle swarm algorithm in the iterative process of "exploration first, then balancing, and then convergence". It can fully explore the feasible solution space in the early stage of iteration and accurately converge to the optimal solution in the later stage. Compared with fixed learning factor or simple linear adjustment, this calculation formula can significantly improve the optimization performance of the algorithm.

[0104] Furthermore, the second learning factor is calculated based on a preset formula; the formula for calculating the second learning factor is as follows: ; Where C2 is the initial value of the second learning factor. This represents the maximum critical value corresponding to the second learning factor. Let be the minimum critical value corresponding to the second learning factor, and d be the ratio. By gradually reducing the base value, the value of C2 can be relatively large in the early to mid-stages of the iteration, making the particles more inclined to explore extensively in the solution space and avoiding premature convergence.

[0105] for sech(4-4d) is a hyperbolic secant function. When d is in the range [0, 0.3] (in the early stages of iteration), the value of sech(4-4d) is relatively small, and it gradually increases as d increases; and, When d is in the range [0, 0.5] (mid-iteration), the corresponding value is relatively small, and it gradually increases as d increases. The product of the two terms can increase the value of this term in the early and mid-iteration stages.

[0106] For example, when d=0.2, sech(3.2) is approximately 0.0002. The product is approximately 0.00006, and C2 is relatively small, making the particles more inclined to explore. When d = 0.5, sech(2) is approximately 0.2659. The product is approximately 0.7071, and the value of this term gradually increases.

[0107] And, as d increases, 0.08(1-d) 2 Gradually decreases, 0.08(1-d) 2 The contribution to C2 gradually decreases. In the later stages of the iteration, when d is in [0.7, 1], the contribution of the previous term is already very small, but this term can still provide a certain contribution.

[0108] It is understandable that, due to the above calculation formula, by The initial value, minus the formulas that gradually increase with the number of iterations ( , The corresponding value, and subtract 0.08(1-d) which still contributes to the increase of the number of iterations. 2 The corresponding numerical value. Therefore, throughout the entire iteration process, due to Since C2 is the initial value, it is relatively large in the early stages of iteration. Furthermore, based on the explanation of C1 above, C1 is relatively small in the early stages of iteration. Therefore, in the early stages of iteration, the particle tends to explore the solution space extensively, avoiding premature convergence to a local optimum. In the middle stages of iteration, as the number of iterations increases, C1 begins to increase, while C2 decreases, and their sizes become similar. This allows the particle to simultaneously focus on its own optimum (pbest) and the global optimum (gbest) in the middle stages of iteration, tending to balance exploration and convergence. Finally, in the later stages of iteration, C1 further increases, while C2 further decreases, thus making the particle more inclined to accelerate convergence to the optimal solution.

[0109] Based on this, dynamically adjusting C1 and C2 using the above calculation formula allows particles to explore more solutions based on their strong global search capabilities in the early stages of particle swarm optimization. Furthermore, in the later stages of the algorithm's iteration, it enhances the particles' reliance on their historical best positions, making them more dependent on their experience and driving them to conduct refined searches within known regions, thereby improving the accuracy of the optimal solution.

[0110] Furthermore, the third learning factor is calculated based on a preset formula; the formula for calculating the third learning factor is as follows: ; Where C3 is the initial value of the third learning factor. This represents the maximum critical value corresponding to the third learning factor. d is the minimum critical value corresponding to the third learning factor, and d is the ratio.

[0111] For the above calculation formula, in the initial stage of iteration (e.g., d is located in [0, 0.3]): It is an exponential function, which decays rapidly when d is in the range [0, 0.3], but the overall value is still relatively large. For example, when d=0, the value of this term is 0.4; when d=0.3, the value of this term is approximately 0.2195. The value of this term is 0.4 when d=0 and approximately 0.2130 when d=0.3. This is a quadratic polynomial, and in the early stages of iteration, the value of this term decreases slowly. For example, when d=0, the value of this term is 0.3; when d=0.3, the value of this term is approximately 0.098.

[0112] Based on the above explanation, it can be concluded that in the early stage of iteration, although the C3 value gradually decreases, the overall value is still relatively large.

[0113] In the middle stage of the iteration (e.g., d is in [0.3, 0.7]) It continues to decay. For example, when d=0.5, the value of this term is 0.1472; when d=0.7, the value of this term is approximately 0.0986. It continues to decay, but the decay rate slows down. When d=0.5, the value of this term is approximately 0.15; when d=0.7, the value of this term is approximately 0.1084. It continues to decay, but the decay rate accelerates. For example, when d=0.5, the value of this term is 0.05; when d=0.7, the value of this term is 0.018.

[0114] Based on the above explanation, it can be assumed that the C3 value continues to decrease gradually during the middle stage of the iteration.

[0115] In the later stages of the iteration (e.g., when d is in [0.7, 1]) The decay continues, and the rate of decay slows down, with the overall value approaching 0. For example, when d=0.9, the value of this term is 0.0661; when d=1, the value of this term is approximately 0.0541. It continues to decay and approaches 0. When d=0.9, the value of this term is approximately 0.0884; when d=0.7, the value of this term is approximately 0.0804. The decay continues, and the decay rate remains relatively fast. For example, when d=0.9, the value of this term is 0.002; when d=1, the value of this term is 0.

[0116] Based on the above explanation, it can be assumed that in the later stages of the iteration, the overall value of C3 approaches 0, that is, the value is relatively small.

[0117] Based on the above analysis, it can be considered that C3 is... As the calculation formula above gradually changes, the C3 value will be larger in the early stages of iteration and smaller in the later stages. Therefore, in the early stages of the algorithm's iteration, a larger C3 value helps particles to explore the solution space more extensively, avoiding getting trapped in local optima. In the middle and later stages of the iteration, to accelerate the convergence of the particle swarm and allow particles to search for the optimal solution more accurately, the C3 value will gradually decrease.

[0118] In addition, it should be noted that by using different formulas to replace different decay rates to reduce the value of C3, the particle swarm optimization algorithm can find a balance between exploration and exploitation, avoid premature convergence, and improve the adaptability and stability of the algorithm.

[0119] Based on the above explanation, in the particle swarm optimization algorithm, precise control of particle search behavior can be achieved at different iteration stages by nonlinearly and dynamically adjusting the three learning factors. In the early iteration stage, the first learning factor C1 gradually increases from a small value, while the second and third learning factors C2 and C3 gradually decrease from a large value. Therefore, in the early iteration stage, particles can perform a broader global search in the solution space, avoiding premature convergence to local regions. In the middle iteration stage, C1 continues to increase, while C2 and C3 continue to decrease. The further increase in C1 strengthens the particle's utilization of its own experience, while the decrease in C2 and C3 brings C2 and C1 closer together, thus allowing the particle to achieve a balance between global search and local optima. Finally, in the later iteration stage, C1 reaches a large value, while C2 and C3 decrease to smaller values. At this point, the large value of C1 makes the particle highly value its historical optimal position, while the decrease in C2 and C3 reduces the particle's utilization of global optimal information. The three factors work together to guide the particle to perform a fine local search within the already discovered optimal region, quickly converging to the global optimum. This avoids meaningless extensive searches in the solution space and significantly improves the convergence accuracy and efficiency of the algorithm.

[0120] S403. Determine the correction coefficient corresponding to the current iteration number based on the position standard deviation.

[0121] In one embodiment, the electronic device can pre-set correction coefficients corresponding to multiple standard deviation ranges, and then determine the correction coefficient corresponding to the standard deviation range in which the position standard deviation is located as the correction coefficient corresponding to the current iteration number.

[0122] In another embodiment, the electronic device can calculate the ratio between the position standard deviation and a preset maximum position standard deviation, and determine the adaptive correction coefficient corresponding to the current iteration number. Then, it calculates the product of the ratio and the adaptive correction coefficient, and finally determines the correction coefficient as the difference between 1 and the product.

[0123] In one embodiment, the aforementioned maximum position standard deviation is a pre-set threshold representing the maximum permissible dispersion of the particle swarm's positions in the solution space. It measures the proportion of the current dispersion of the particle swarm's positions (quantified by the position standard deviation) relative to the maximum possible dispersion. It is understood that if the position standard deviation of the particle swarm exceeds the preset maximum position standard deviation, it may indicate that the particle swarm is too dispersed; conversely, if the position standard deviation is much smaller than the maximum position standard deviation, it may indicate that the particle swarm is already relatively concentrated, and the solution process will focus more on local search.

[0124] The above ratio reflects the current dispersion of the particle swarm relative to the maximum permissible dispersion.

[0125] The aforementioned adaptive correction coefficient is a dynamically determined coefficient based on the current iteration number, used to adjust the correction strength according to the iteration progress. It can be determined based on a preset mapping relationship between the iteration number and the coefficient.

[0126] For example, in the early stages of iteration, the adaptive correction coefficient is larger to enhance the adjustment of the particle swarm dispersion, giving the algorithm a stronger exploratory capability; in the later stages of iteration, the adaptive correction coefficient is smaller to avoid over-adjustment that could cause the particle swarm to converge prematurely, giving the algorithm a stronger expansive capability. This dynamic adjustment method allows the algorithm to maintain good search performance in different iteration stages.

[0127] In this embodiment, by calculating the aforementioned ratio, the dispersion of the particle swarm can be normalized, reflecting the proportion of the current dispersion relative to the maximum allowable dispersion. Then, the adaptive correction coefficient is dynamically determined based on the current iteration count, allowing the correction strength to adapt to the needs of the algorithm at different iteration stages. Next, multiplying the ratio by the adaptive correction coefficient yields an intermediate result that comprehensively considers both the particle swarm dispersion and the iteration progress. Finally, the difference between 1 and the product is determined as the correction coefficient, which can be used to adjust the aforementioned learning factor. Furthermore, the magnitude of the correction coefficient reflects the degree of adjustment required for the learning factor at the current iteration count, thereby achieving a dynamic balance between the algorithm's exploration and development capabilities.

[0128] S404. The product of the correction coefficient and the initial value is determined as the target value.

[0129] In one embodiment, calculating the ratio of the current iteration number to the preset total iteration number can intuitively quantify the iteration progress in solving the objective function. Then, an initial value is determined based on the critical value difference within the preset range of the learning factor and the aforementioned ratio. This initial value, adapted to the current stage, can be generated by combining the difference and ratio between the maximum and minimum critical values, thus achieving phased basic control of the learning factor. Next, a correction coefficient corresponding to the current iteration number is determined based on the position standard deviation, and the product of the correction coefficient and the initial value is used to determine the target value. This introduces the correction coefficient into the "real-time distribution state" of the particle swarm (reflected by the position standard deviation), compensating for the mechanical nature of setting the initial value solely based on the iteration progress. This allows the target value to comprehensively consider the macroscopic needs of the iteration stage and the real-time state of the particle distribution, achieving adaptability to different iteration stages.

[0130] In another embodiment, the third learning factor C3 is used in the particle swarm optimization algorithm to balance the local and global searches of particles. Therefore, in the early stages of iteration, a larger target value helps particles explore the solution space more broadly, avoiding getting trapped in local optima. In the later stages of iteration, to accelerate the convergence of the particle swarm and allow particles to search for optimal solutions more accurately, C3 is typically gradually decreased. Also, in the early stages of iteration, particles need strong global search capabilities, so the target values ​​of C1 and C2 can be relatively balanced (e.g., 1.5). However, in the later stages of iteration, to better utilize the experience of the particle swarm and accelerate the discovery of local optima, C1 and C2 can be adjusted. For example, in the later stages of iteration, increasing C1 enhances the particle's dependence on its historical optimal position, driving it to perform a refined search within known regions. Simultaneously, decreasing C2 reduces the particle's dependence on the global optimal position, thereby improving local search accuracy.

[0131] S303. Update the particle corresponding to the current iteration number according to the target value to obtain the iterated particle.

[0132] In this embodiment, the particles obtained through the aforementioned iterations serve as the input to the objective function. In one embodiment, after obtaining the objective value, it can be substituted into the aforementioned velocity update formula to obtain the updated velocity, and the position can be updated based on the updated velocity to obtain the iterated particles. This process will not be described in detail here.

[0133] In this embodiment, for any iteration in the iterative phase, the current iteration number and the corresponding particle position standard deviation are obtained. Based on the position standard deviation and the current iteration number, the target value of the learning factor is determined. This allows the adjustment of the learning factor (target value) to meet the phase requirements, resolving the imbalance between exploration and development caused by traditional fixed-learning-factor iterations. Finally, the particles are updated according to the target value and input into the objective function to obtain the charging and discharging power at the current iteration number, providing a basis for subsequent calculations of the target charging and discharging power.

[0134] Please see Figure 5 , Figure 5 This is a schematic diagram of a hybrid energy storage optimization scheduling device provided in an embodiment of this application. The modules included in the hybrid energy storage optimization scheduling device in this embodiment are used to perform… Figures 1 to 4 The steps in the corresponding embodiments. Please refer to the details. Figures 1 to 4 The relevant descriptions in the corresponding embodiments are shown below. For ease of explanation, only the parts relevant to this embodiment are shown. See also... Figure 5 The hybrid energy storage optimization scheduling device 500 may include: a first acquisition module 510, a construction module 520, a second acquisition module 530, an establishment module 540, a solution module 550, and a scheduling module 560, wherein: The first acquisition module 510 is used to acquire the power parameters of the hybrid energy storage system and the grid operation parameters of the power grid; the power parameters are used to describe the physical characteristics of the hybrid energy storage system; the grid operation parameters include the maximum safe power during grid operation and the grid time-of-use electricity price.

[0135] Module 520 is used to construct the operating constraints of the hybrid energy storage system based on power parameters and maximum safe power.

[0136] The second acquisition module 530 is used to acquire a first sequence of power generation forecast data corresponding to each future time point of the new energy power generation equipment, and a second sequence of load forecast data.

[0137] Module 540 is established to create an objective function based on the first sequence, the second sequence, and the grid time-of-use electricity price, with the charging and discharging power of the hybrid energy storage system at each future time point as the unknown quantity and the charging and discharging cost as the optimization objective.

[0138] The solver module 550 is used to solve the objective function using the particle swarm optimization algorithm to obtain the target charge and discharge power of the hybrid energy storage system at each future time point; during the process of solving the objective function, the charge and discharge power and the change in the amount of electricity of the hybrid energy storage system satisfy the operating constraints.

[0139] The scheduling module 560 is used to schedule the charging and discharging of the hybrid energy storage system based on the target charging and discharging power at various future time points.

[0140] In one embodiment, the solver module 550 is further configured to: In the particle swarm initialization phase and the iterative update phase for optimizing particles by solving the objective function, if a particle violates the operating constraints, the particle is corrected using a preset correction rule; the corrected particle satisfies the operating constraints; the particle represents a charging and discharging power at each future time point; the charging and discharging power corresponding to the particle after the end of the iteration phase is determined as the target charging and discharging power.

[0141] In one embodiment, the solver module 550 is further configured to: For any iteration in the iteration phase, obtain the current iteration number and the position standard deviation of the particle corresponding to the current iteration number; based on the position standard deviation and the current iteration number, determine the target value of the learning factor corresponding to the current iteration number; update the particle corresponding to the current iteration number according to the target value to obtain the iterated particle; the iterated particle is used as the input of the objective function.

[0142] In one embodiment, the solver module 550 is further configured to: Calculate the ratio of the current iteration number to the preset total iteration number; determine the initial value of the learning factor corresponding to the current iteration number based on the difference and ratio between the maximum and minimum critical values ​​in the preset range corresponding to the learning factor; determine the correction coefficient corresponding to the current iteration number based on the position standard deviation; and determine the target value by multiplying the correction coefficient and the initial value.

[0143] In one embodiment, the learning factor is divided into a first learning factor for the particle to perform a local search of its own historical best position, a second learning factor for the particle to perform a global search of its global best position, and a third learning factor for balancing the particle's local and global searches; the solution module 550 is also used for: The initial value of the first learning factor is calculated based on a preset formula; the formula for calculating the first learning factor is as follows: ; Where C1 is the initial value of the first learning factor. The maximum critical value corresponding to the first learning factor. is the minimum critical value corresponding to the first learning factor, and d is the ratio; The second learning factor is calculated based on a preset formula; the formula for calculating the second learning factor is as follows: ; Where C2 is the initial value of the second learning factor. This represents the maximum critical value corresponding to the second learning factor. d is the minimum critical value corresponding to the second learning factor, and d is the ratio. The third learning factor is calculated based on a preset formula; the formula for calculating the third learning factor is as follows: ; Where C3 is the initial value of the third learning factor. This represents the maximum critical value corresponding to the third learning factor. d is the minimum critical value corresponding to the third learning factor, and d is the ratio.

[0144] In one embodiment, the solver module 550 is further configured to: Calculate the ratio between the position standard deviation and the preset maximum position standard deviation; determine the adaptive correction coefficient corresponding to the current iteration number; calculate the product of the ratio and the adaptive correction coefficient; and determine the difference between 1 and the product as the correction coefficient.

[0145] In one embodiment, the hybrid energy storage system includes two types of energy storage devices, and the objective function is: ; in, For charging and discharging costs, Indicates the cost of grid interaction. This indicates the switching cost of energy storage devices during charge / discharge cycles. Let T be the equipment lifespan cost of the energy storage device, T be the total number of future time points, and t be the t-th future time point. This is the predicted power generation data for the t-th future time point. For the load forecast data at the t-th future time point, Let be the charging and discharging power of the first type of energy storage device at the future time point t. Let be the charging and discharging power of the second type of energy storage device at the t-th future time point, and step be the preset time period corresponding to two adjacent future time points. Let t be the time-of-use electricity price for the power grid at the t-th future time point. Let be the number of charge-discharge reversals of the first type of energy storage device at a future time point t. Let $\frac{ ... Let be the number of charge-discharge reversals for the second type of energy storage device at the future time point t. Let $\frac{ ... These represent preset coefficients. This represents the health status of the first type of energy storage device at a future time point t. This is the reference health status for the first type of energy storage device. The health status of the second type of energy storage device at a future time point t. This is a reference health status for the second type of energy storage device.

[0146] In one embodiment, the first energy storage device is a lithium iron phosphate energy storage device, and the second energy storage device is a vanadium redox flow storage device. The preset coefficient corresponding to the lithium iron phosphate energy storage device is greater than the preset coefficient corresponding to the vanadium redox flow storage device, and the initial switching cost corresponding to the lithium iron phosphate energy storage device is greater than the initial switching cost corresponding to the vanadium redox flow storage device.

[0147] When it is understood that, Figure 5 In the schematic diagram of the hybrid energy storage optimized scheduling device shown, each module is used to perform... Figures 1 to 4 The steps in the corresponding embodiments, and for Figures 1 to 4 The steps in the corresponding embodiments have been explained in detail in the above embodiments. Please refer to them for details. Figures 1 to 4 The relevant descriptions in the corresponding embodiments will not be repeated here.

[0148] Figure 6 This is a schematic diagram of the structure of an electronic device provided in one embodiment of this application. Figure 6As shown, the electronic device 600 of this embodiment includes: a processor 610, a memory 620, and a computer program 630 stored in the memory 620 and executable by the processor 610, such as a program for a hybrid energy storage optimization scheduling method. When the processor 610 executes the computer program 630, it implements the steps in the various embodiments of the hybrid energy storage optimization scheduling method described above, for example... Figure 1 S101 to S106 are shown. Alternatively, the processor 610 implements the above when executing the computer program 630. Figure 5 The functions of each module in the corresponding embodiments, for example, Figure 5 For details on the functions of each module shown, please refer to [link / reference]. Figure 5 The relevant descriptions in the corresponding embodiments.

[0149] For example, the computer program 630 can be divided into one or more modules, one or more of which are stored in the memory 620 and executed by the processor 610 to implement the hybrid energy storage optimization scheduling method provided in the embodiments of this application. One or more modules can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program 630 in the electronic device 600. For example, the computer program 630 can implement the hybrid energy storage optimization scheduling method provided in the embodiments of this application.

[0150] Electronic device 600 may include, but is not limited to, processor 610 and memory 620. Those skilled in the art will understand that... Figure 6 This is merely an example of electronic device 600 and does not constitute a limitation on electronic device 600. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device may also include input / output devices, network access devices, buses, etc.

[0151] The processor 610 may be a central processing unit, or it may be other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0152] The memory 620 can be an internal storage unit of the electronic device 600, such as a hard disk or memory of the electronic device 600. The memory 620 can also be an external storage device of the electronic device 600, such as a plug-in hard disk, smart memory card, flash memory card, etc., equipped on the electronic device 600. Furthermore, the memory 620 can include both internal storage units and external storage devices of the electronic device 600.

[0153] This application provides a computer-readable storage medium, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the hybrid energy storage optimization scheduling method as described in the above embodiments.

[0154] This application provides a computer program product that, when run on an electronic device, causes the electronic device to execute the hybrid energy storage optimization scheduling method described in the above embodiments.

[0155] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A hybrid energy storage optimal scheduling method, characterized in that, The method comprises: acquiring power parameters of a hybrid energy storage system and grid operation parameters of a power grid; the power parameters are used to describe physical characteristics of the hybrid energy storage system; the grid operation parameters include maximum safe power when the power grid is running and time-of-use electricity price of the power grid; constructing operation constraint conditions of the hybrid energy storage system according to the power parameters and the maximum safe power; acquiring a first sequence composed of power generation prediction data corresponding to each future time point respectively of a new energy power generation device and a second sequence composed of each load prediction data; based on the first sequence, the second sequence and the time-of-use electricity price, establishing a target function taking the charging and discharging power of the hybrid energy storage system at the future time points as unknown quantities and taking the charging and discharging cost as an optimization target; solving the target function by using a particle swarm algorithm to obtain the target charging and discharging power of the hybrid energy storage system at the future time points; in the process of solving the target function, the charging and discharging power of the hybrid energy storage system and the power change satisfy the operation constraint conditions; based on the target charging and discharging power, scheduling the charging and discharging of the hybrid energy storage system at the future time points.

2. The method of claim 1, wherein, The method comprises: in the particle swarm initialization stage and the iterative updating stage of optimizing the particles in solving the target function, if the particles violate the operation constraint conditions, the particles are corrected by using a preset correction rule; the corrected particles satisfy the operation constraint conditions; the particles represent a charging and discharging power at the future time points; the charging and discharging power corresponding to the particles after the iterative stage is ended is determined as the target charging and discharging power.

3. The method of claim 2, wherein, Before the charging and discharging power corresponding to the particles after the iterative stage is ended is determined as the target charging and discharging power, the method further comprises: for any iteration process in the iterative stage, acquiring the current iteration number and the position standard deviation of the particles corresponding to the current iteration number; based on the position standard deviation and the current iteration number, determining a target value of a learning factor corresponding to the current iteration number; updating the particles corresponding to the current iteration number according to the target value to obtain the particles after iteration; the particles after iteration are used as the input of the target function.

4. The method of claim 3, wherein, The method comprises: calculating the ratio of the current iteration number to a preset total iteration number; based on the difference between the maximum critical value and the minimum critical value in the preset range corresponding to the learning factor and the ratio, determining an initial value of the learning factor corresponding to the current iteration number; based on the position standard deviation, determining a correction coefficient corresponding to the current iteration number; determining the product of the correction coefficient and the initial value as the target value.

5. The method of claim 4, wherein, The learning factor is divided into a first learning factor for the particle to locally search for the optimal position of its own history, a second learning factor for the particle to globally search for the global optimal position, and a third learning factor for balancing the particle to perform the local search and the global search; and the initial value of the learning factor corresponding to the current iteration number is determined based on the difference between the maximum critical value and the minimum critical value in the preset range corresponding to the learning factor and the ratio, and includes: The initial value of the first learning factor is calculated based on a preset first learning factor calculation formula; the first learning factor calculation formula is as follows: ； C1 is an initial value of the first learning factor, Cmax is a maximum critical value corresponding to the first learning factor, Cmin is a minimum critical value corresponding to the first learning factor, and d is the ratio. The second learning factor is calculated based on a preset second learning factor calculation formula; the second learning factor calculation formula is as follows: ； C2 is an initial value of the second learning factor, Cmax is a maximum critical value corresponding to the second learning factor, Cmin is a minimum critical value corresponding to the second learning factor, and d is the ratio. The third learning factor is calculated based on a preset third learning factor calculation formula; the third learning factor calculation formula is as follows: ； C3 is an initial value of the third learning factor, C4 is a maximum critical value corresponding to the third learning factor, C5 is a minimum critical value corresponding to the third learning factor, and d is the ratio.

6. The method of claim 4, wherein, The correction coefficient corresponding to the current iteration number is determined based on the position standard deviation, including: The ratio between the position standard deviation and a preset maximum position standard deviation is calculated; An adaptive correction coefficient corresponding to the current iteration number is determined; The product of the ratio and the adaptive correction coefficient is calculated; The difference between 1 and the product is determined as the correction coefficient.

7. The method according to any one of claims 1 to 6, characterized in that, The hybrid energy storage system includes two energy storage devices, and the target function is: ； wherein, is the charging and discharging cost of the energy storage device, represents the grid interaction cost, represents the flip cost of the charging and discharging flip of the energy storage device, is the device life cost of the energy storage device, T is the total number of time points of the future time points, and t is the tth future time point, is the power generation prediction data of the tth future time point, is the load prediction data of the tth future time point, is the charging and discharging power of the first energy storage device at the tth future time point, is the charging and discharging power of the second energy storage device at the tth future time point, and step is a preset time period corresponding to adjacent two future time points, is the time-of-use price of the grid at the tth future time point, is the number of charging and discharging flips of the first energy storage device at the tth future time point, is the initial flip cost of the first energy storage device corresponding to the tth future time point, is the number of charging and discharging flips of the second energy storage device corresponding to the tth future time point, is the initial flip cost of the second energy storage device corresponding to the tth future time point, respectively represent preset coefficients, represents the health state of the first energy storage device at the tth future time point, is the reference health state of the first energy storage device, represents the health state of the second energy storage device at the tth future time point, is the reference health state of the second energy storage device.

8. The method of claim 7, wherein, The first energy storage device is a lithium iron phosphate energy storage device, and the second energy storage device is a full vanadium flow energy storage device; the preset coefficient corresponding to the lithium iron phosphate energy storage device is greater than the preset coefficient corresponding to the full vanadium flow energy storage device, and the initial flip cost corresponding to the lithium iron phosphate energy storage device is greater than the initial flip cost corresponding to the full vanadium flow energy storage device.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the method of any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, When the computer program product is running on the electronic device, the electronic device is caused to execute to realize the method of any one of claims 1-8. When the computer program product is running on the electronic device, the electronic device is caused to execute to realize the method of any one of claims 1-8.

Images