Carbon emission prediction and model training method and device, equipment, storage medium and program product

Abstract

A carbon emission prediction and model training method and device, equipment, storage medium and program product are disclosed, relating to the technical field of carbon metering. The method comprises: obtaining operation data of a battery energy storage system; determining cycle attenuation and calendar attenuation based on the operation data; determining a dynamic coupling coefficient based on the cycle attenuation and the calendar attenuation; determining a cross-coupling term based on the product of the dynamic coupling coefficient, the cycle attenuation and the calendar attenuation; predicting energy efficiency based on the cycle attenuation, the calendar attenuation and the cross-coupling term; and predicting carbon emission based on the energy efficiency. The above method first determines the cycle attenuation and the calendar attenuation based on the operation data, then determines the dynamic coupling coefficient based on the cycle attenuation and the calendar attenuation, and further determines the cross-coupling term, thereby predicting the energy efficiency and the carbon emission, and improving the accuracy of the prediction of the energy efficiency and the carbon emission.

Description

Technical Field

[0001] This application relates to the field of carbon measurement technology, and in particular to a method, apparatus, equipment, storage medium and program product for predicting and training carbon emissions. Background Technology

[0002] In related technologies, the carbon emissions of battery energy storage systems are typically predicted based on the charging and discharging capacity and fixed loss rate of the battery energy storage system. For example, the charging and discharging capacity of the battery energy storage system in a certain cycle can be multiplied by the fixed loss rate to obtain the lost capacity, and then multiplied by a preset electricity carbon emission factor to obtain the corresponding carbon emissions.

[0003] However, the above method of using a fixed loss rate to predict carbon emissions leads to inaccurate predictions of carbon emissions. Summary of the Invention

[0004] This application provides a method, apparatus, device, storage medium, and program product for predicting and training carbon emissions. The technical solution provided by this application includes the following aspects.

[0005] According to one aspect of the embodiments of this application, a method for predicting carbon emissions is provided, the method comprising: Obtain the operating data of the battery energy storage system in the i-th cycle, where i is a positive integer; Based on the operating data within the i-th cycle, determine the cyclic decay amount for the i-th cycle, and based on the operating data within the i-th cycle, determine the calendar decay amount for the i-th cycle. Based on the cyclic decay amount and the calendar decay amount of the i-th cycle, the dynamic coupling coefficient of the i-th cycle is determined. The dynamic coupling coefficient is used to indicate the coupling strength between cyclic aging and calendar aging within the i-th cycle. The dynamic coupling coefficient is positively correlated with the cyclic decay amount and the calendar decay amount. Based on the product of the dynamic coupling coefficient of the i-th cycle, the cyclic decay of the i-th cycle, and the calendar decay of the i-th cycle, the cross-coupling term of the i-th cycle is determined. The cross-coupling term is used to indicate the positive feedback coupling effect between cyclic aging and calendar aging that increases with the degree of aging. Based on the cyclic decay amount of the i-th period, the calendar decay amount of the i-th period, and the cross-coupling term of the i-th period, predict the energy efficiency corresponding to the i-th period; Based on the energy efficiency corresponding to the i-th cycle, predict the carbon emissions corresponding to the i-th cycle.

[0006] According to one aspect of the embodiments of this application, a model training method is provided, the method comprising: Obtain the operating data of the battery energy storage system in the i-th cycle, where i is a positive integer; Based on the running data within the i-th period, the cyclic decay amount and calendar decay amount of the i-th period are determined using a multi-factor decay prediction model. The multi-factor decay prediction model determines the dynamic coupling coefficient of the i-th period based on the cyclic decay amount and the calendar decay amount of the i-th period. The dynamic coupling coefficient is used to indicate the coupling strength between cyclic aging and calendar aging within the i-th period. The dynamic coupling coefficient is positively correlated with the cyclic decay amount and the calendar decay amount. The multi-factor decay prediction model determines the cross-coupling term for the i-th period based on the product of the dynamic coupling coefficient, the cyclic decay amount, and the calendar decay amount for the i-th period. The cross-coupling term is used to indicate the positive feedback coupling effect between cyclic aging and calendar aging that increases with the degree of aging. The multi-factor decay prediction model predicts the energy efficiency corresponding to the i-th cycle based on the cyclic decay amount of the i-th cycle, the calendar decay amount of the i-th cycle, and the cross-coupling term of the i-th cycle. Based on the energy efficiency corresponding to the i-th cycle, predict the carbon emissions corresponding to the i-th cycle; If the prediction deviation meets the preset conditions, the correctable parameters in the multi-factor decay prediction model are adjusted based on the prediction deviation to obtain the trained multi-factor decay prediction model; wherein, the prediction deviation is determined based on the deviation between the carbon emissions corresponding to the i-th period and the actual carbon emissions corresponding to the i-th period.

[0007] According to one aspect of the embodiments of this application, a carbon emission prediction device is provided, the device comprising: The acquisition module is used to acquire the operating data of the battery energy storage system in the i-th cycle, where i is a positive integer; The first determining module is used to determine the cyclic decay amount of the i-th period based on the running data within the i-th period, and to determine the calendar decay amount of the i-th period based on the running data within the i-th period. The second determining module is used to determine the dynamic coupling coefficient of the i-th period based on the cyclic decay amount and the calendar decay amount of the i-th period. The dynamic coupling coefficient is used to indicate the coupling strength between cyclic aging and calendar aging in the i-th period. The dynamic coupling coefficient is positively correlated with the cyclic decay amount and the calendar decay amount. The third determining module is used to determine the cross-coupling term of the i-th period based on the product of the dynamic coupling coefficient of the i-th period, the cyclic decay amount of the i-th period, and the calendar decay amount of the i-th period. The cross-coupling term is used to indicate the positive feedback coupling effect between cyclic aging and calendar aging that increases with the degree of aging. The first prediction module is used to predict the energy efficiency corresponding to the i-th period based on the cyclic decay amount of the i-th period, the calendar decay amount of the i-th period, and the cross-coupling term of the i-th period. The second prediction module is used to predict the carbon emissions corresponding to the i-th cycle based on the energy efficiency corresponding to the i-th cycle.

[0008] According to one aspect of the embodiments of this application, a model training apparatus is provided, the apparatus comprising: The acquisition module is used to acquire the operating data of the battery energy storage system in the i-th cycle, where i is a positive integer; The first determining module is used to determine the cyclic decay amount and the calendar decay amount of the i-th period based on the running data in the i-th period using a multi-factor decay prediction model. The second determining module is used to determine the dynamic coupling coefficient of the i-th period based on the cyclic decay amount and the calendar decay amount of the i-th period through the multi-factor decay prediction model. The dynamic coupling coefficient is used to indicate the coupling strength between cyclic aging and calendar aging in the i-th period. The dynamic coupling coefficient is positively correlated with the cyclic decay amount and the calendar decay amount. The third determining module is used to determine the cross-coupling term of the i-th period based on the product of the dynamic coupling coefficient of the i-th period, the cyclic decay amount of the i-th period, and the calendar decay amount of the i-th period through the multi-factor decay prediction model. The cross-coupling term is used to indicate the positive feedback coupling effect between cyclic aging and calendar aging that increases with the degree of aging. The first prediction module is used to predict the energy efficiency corresponding to the i-th period based on the cyclic decay amount of the i-th period, the calendar decay amount of the i-th period, and the cross-coupling term of the i-th period through the multi-factor decay prediction model. The second prediction module is used to predict the carbon emissions corresponding to the i-th cycle based on the energy efficiency corresponding to the i-th cycle. The module is used to adjust the correctable parameters in the multi-factor decay prediction model based on the prediction deviation when the prediction deviation meets the preset conditions, so as to obtain the trained multi-factor decay prediction model; wherein the prediction deviation is determined based on the deviation between the carbon emissions corresponding to the i-th period and the actual carbon emissions corresponding to the i-th period.

[0009] According to one aspect of the embodiments of this application, a computer device is provided, the computer device including a processor and a memory, the memory storing a computer program, the computer program being loaded and executed by the processor to implement the above-described carbon emission prediction method, or the above-described model training method.

[0010] According to one aspect of the embodiments of this application, a computer-readable storage medium is provided, wherein a computer program is stored in the computer-readable storage medium, the computer program being loaded and executed by a processor to implement the above-described carbon emission prediction method or the above-described model training method.

[0011] According to one aspect of the embodiments of this application, a computer program product is provided, the computer program product including computer instructions, which are executed by a processor to implement the above-described carbon emission prediction method or the above-described model training method.

[0012] The technical solution provided in this application can bring the following beneficial effects: This application does not directly predict the carbon emissions of a battery energy storage system using a fixed loss rate. Instead, it first determines the cycle decay and calendar decay for the i-th cycle based on the operating data within the i-th cycle. Then, it determines the dynamic coupling coefficient based on the cycle decay and calendar decay, and further determines the cross-coupling term, thereby predicting the energy efficiency and carbon emissions corresponding to the i-th cycle. This avoids the inaccurate prediction problem caused by using a fixed loss rate to predict carbon emissions in related technologies, and improves the accuracy of energy efficiency and carbon emissions.

[0013] Furthermore, characterizing the coupling strength between cyclic aging and calendar aging using a dynamic coupling coefficient allows the coupling strength to increase with the degree of battery aging, avoiding the problem of difficulty in adapting to different aging stages when using a fixed coupling coefficient, thereby improving the accuracy of battery aging state characterization. Simultaneously, characterizing the positive feedback coupling effect between cyclic aging and calendar aging, which increases with the degree of aging, using a cross-coupling term, can further reflect the mutually reinforcing effect between the two aging mechanisms. This allows for the introduction of the additional degradation effect caused by the combined effect of cyclic aging and calendar aging in the energy efficiency prediction process, thereby further improving the accuracy of energy efficiency and carbon emission prediction results. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of a computer system provided in one embodiment of this application; Figure 2 This is a flowchart of a carbon emission prediction method provided in one embodiment of this application; Figure 3 This is a flowchart of a model training method provided in one embodiment of this application; Figure 4 This is a flowchart of a model training method provided in another embodiment of this application; Figure 5 This is a block diagram of a carbon emission prediction device provided in one embodiment of this application; Figure 6 This is a block diagram of a model training apparatus provided in one embodiment of this application; Figure 7 This is a structural block diagram of a computer device provided in one embodiment of this application. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0016] Please refer to Figure 1 The diagram illustrates a computer system provided in one embodiment of this application. The computer system may include a model training device 110 and a model usage device 120.

[0017] The model training device 110 can be an electronic device such as a mobile phone, desktop computer, tablet computer, laptop computer, vehicle terminal, server, intelligent robot, smart TV, multimedia playback device, or other electronic devices with strong computing power; this application does not limit this. The model training device 110 is used to train a multi-factor decay prediction model.

[0018] In this embodiment, the multi-factor decay prediction model can be a machine learning model, or an empirical model, semi-empirical model, or physical mechanism model containing calibrable parameters. Optionally, the model training device 110 can train the multi-factor decay prediction model using machine learning, parameter fitting, or parameter optimization to achieve better prediction performance. Optionally, the training process of the multi-factor decay prediction model is as follows (this is only a brief description; the specific training process is described in the following embodiments): acquire the operating data of the battery energy storage system in the i-th cycle; determine the cyclic decay amount and calendar decay amount of the i-th cycle based on the operating data in the i-th cycle using the multi-factor decay prediction model; determine the dynamic coupling coefficient of the i-th cycle based on the cyclic decay amount and calendar decay amount of the i-th cycle using the multi-factor decay prediction model; determine the dynamic coupling coefficient of the i-th cycle based on the dynamic coupling coefficient and calendar decay amount of the i-th cycle using the multi-factor decay prediction model; and determine the cyclic decay amount and calendar decay amount of the i-th cycle using the multi-factor decay prediction model based on the dynamic coupling coefficient and calendar decay amount of the i-th cycle. The cross-coupling term for the i-th period is determined by multiplying the cyclic decay amount of the i-th period and the calendar decay amount of the i-th period. The energy efficiency corresponding to the i-th period is predicted using a multi-factor decay prediction model based on the cyclic decay amount, calendar decay amount, and cross-coupling term of the i-th period. Based on the energy efficiency of the i-th period, the carbon emissions corresponding to the i-th period are predicted. If the prediction deviation meets preset conditions, the correctable parameters in the multi-factor decay prediction model are adjusted based on the prediction deviation to obtain the trained multi-factor decay prediction model.

[0019] The device 120 for model use can be an electronic device such as a mobile phone, desktop computer, tablet computer, laptop computer, vehicle terminal, server, intelligent robot, smart TV, multimedia playback device, or other electronic devices with strong computing power; this application does not limit this. The device 120 for model use can employ a trained multi-factor decay prediction model to predict the energy efficiency corresponding to the i-th period, and predict the carbon emissions corresponding to the i-th period based on the energy efficiency.

[0020] The model training device 110 and the model usage device 120 can be two separate devices or the same device.

[0021] The method provided in this application embodiment can be executed by a computer device, which refers to an electronic device with data computing, processing, and storage capabilities. When the computer device is a server, the server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The computer device can be... Figure 1 The model training device 110 can also be the model usage device 120.

[0022] Please refer to Figure 2 The diagram illustrates a flowchart of a carbon emission prediction method provided in one embodiment of this application. The execution entity for each step of the method can be a computer device; for example, the computer device could be... Figure 1 The model in the computer system shown uses device 120. The method may include at least one of the following steps (210-260).

[0023] Step 210: Obtain the operating data of the battery energy storage system in the i-th cycle, where i is a positive integer.

[0024] A battery energy storage system is an energy storage system that uses an electrochemical battery as an energy storage medium and is able to store electrical energy during charging and release electrical energy during discharging.

[0025] In this application, the i-th cycle refers to a time unit for performing operational analysis, aging assessment, energy efficiency prediction, and carbon emission prediction of the battery energy storage system. The i-th cycle can be a preset time period, such as a day, a week, a month, a scheduling cycle, a metering cycle, or an operation and maintenance analysis cycle; this application does not limit this. For example, when carbon emission prediction is performed on a daily basis, the i-th cycle can represent the i-th day; when carbon emission prediction is performed on a scheduling cycle basis, the i-th cycle can represent the i-th charge and discharge scheduling cycle; when prediction is performed on an operation and maintenance cycle basis, the i-th cycle can represent the i-th operation and maintenance statistics cycle.

[0026] The operational data in the i-th cycle refers to the data generated, collected, statistically analyzed, or calculated during the operation of the battery energy storage system in the i-th cycle, used to characterize the operating status, charging and discharging behavior, thermal state, SOC (State of Charge), and capacity state of the battery energy storage system. This operational data can be collected or generated by at least one of the following: BMS (Battery Management System), PCS (Power Conversion System), EMS (Energy Management System), thermal management system, metering device, sensor, edge controller, and cloud-based operation and maintenance platform; this application does not limit the scope of the data.

[0027] Step 220: Based on the running data in the i-th cycle, determine the cyclic decay amount in the i-th cycle, and based on the running data in the i-th cycle, determine the calendar decay amount in the i-th cycle.

[0028] Cycle decay characterizes the degree of battery aging caused by charge-discharge cycles in the i-th cycle of a battery energy storage system. Calendar decay characterizes the degree of battery aging caused by factors such as time, temperature, and state of charge residence in the i-th cycle of a battery energy storage system.

[0029] During the operation of a battery energy storage system, battery aging typically includes two types of factors: cycle aging and calendar aging. Cycle aging is mainly related to the battery's charge and discharge behavior, such as the number of cycles, depth of discharge, charge / discharge rate, SOC fluctuation, and temperature during cycling. Calendar aging is mainly related to the battery's natural degradation over time, such as operating time, resting time, average temperature, and prolonged residence in different SOC ranges. For details on how to determine cycle aging and calendar aging, please see below.

[0030] Step 230: Based on the cyclic decay amount and calendar decay amount of the i-th cycle, determine the dynamic coupling coefficient of the i-th cycle. The dynamic coupling coefficient is used to indicate the coupling strength between cyclic aging and calendar aging in the i-th cycle. The dynamic coupling coefficient is positively correlated with the cyclic decay amount and the calendar decay amount.

[0031] In the actual operation of battery energy storage systems, cycle aging and calendar aging are not always independent of each other. For example, prolonged exposure to high temperatures or high SOC may alter the state of battery materials, thereby accelerating degradation during subsequent cycles; frequent deep cycling may also accelerate the increase in battery internal resistance or the loss of active lithium, thus affecting calendar degradation during subsequent rest periods. Therefore, this application describes the coupling strength between cycle aging and calendar aging using a dynamic coupling coefficient.

[0032] The greater the cycle decay and calendar decay, the stronger the effects of cycle aging and calendar aging on the battery energy storage system in the i-th cycle, and the more significant the degradation of the battery's internal material state, internal resistance state, active lithium loss, or usable capacity. In this case, the mutually reinforcing effect between cycle aging and calendar aging is also more pronounced. Therefore, this application sets the dynamic coupling coefficient to be positively correlated with both cycle decay and calendar decay.

[0033] The aforementioned positive correlation can be understood as follows: Under otherwise unchanged conditions, when the cyclic decay in the i-th cycle increases, the dynamic coupling coefficient either increases or does not decrease; when the calendar decay in the i-th cycle increases, the dynamic coupling coefficient either increases or does not decrease; when both cyclic decay and calendar decay increase simultaneously, the dynamic coupling coefficient further increases. Therefore, the dynamic coupling coefficient can dynamically increase with the deepening of battery aging, thus reflecting the gradually strengthening coupling effect between cyclic aging and calendar aging.

[0034] In some embodiments, determining the dynamic coupling coefficient of the i-th period based on the cyclic decay amount and the calendar decay amount of the i-th period includes: determining the comprehensive decay amount based on the sum of the cyclic decay amount and the calendar decay amount of the i-th period; and determining the dynamic coupling coefficient of the i-th period based on the product between the coupling acceleration factor and the comprehensive decay amount, wherein the coupling acceleration factor is used to indicate the degree of enhancement effect of the comprehensive decay amount on the dynamic coupling coefficient.

[0035] In some embodiments, the sum of the cyclic decay amount and the calendar decay amount of the i-th period is determined as the total decay amount. For example, the cyclic decay amount of the i-th period can be denoted as D_cyc(i), the calendar decay amount of the i-th period can be denoted as D_cal(i), and the total decay amount of the i-th period can be denoted as D_sum(i). Then the total decay amount can be expressed as: D_sum(i) = D_cyc(i) + D_cal(i).

[0036] In some embodiments, the dynamic coupling coefficient for the i-th cycle is determined based on the product of the coupling acceleration factor and the overall degradation amount, as well as the basic coupling coefficient. The basic coupling coefficient indicates the fundamental coupling strength between cyclic aging and calendar aging when the battery storage system is in a baseline aging state or an initial aging state.

[0037] For example, the dynamic coupling coefficient of the i-th cycle can be denoted as λ(i), the basic coupling coefficient as λ0, and the coupling acceleration factor as φ. Then, the dynamic coupling coefficient of the i-th cycle can be expressed as: λ(i) = λ0 × (1 + φ × D_sum(i)). Substituting D_sum(i) = D_cyc(i) + D_cal(i) into the equation, we get: λ(i) = λ0 × [1 + φ × (D_cyc(i) + D_cal(i))]. From this formula, we can see that when λ0 and φ are positive, the dynamic coupling coefficient λ(i) is positively correlated with the overall attenuation D_sum(i). Since D_sum(i) is obtained by adding the cyclic attenuation D_cyc(i) and the calendar attenuation D_cal(i), the dynamic coupling coefficient λ(i) is also positively correlated with both the cyclic attenuation and the calendar attenuation. In other words, under the condition that other factors remain unchanged, when the cyclic decay amount increases, the calendar decay amount increases, or both increase simultaneously, the dynamic coupling coefficient increases accordingly.

[0038] The above method determines the overall degradation amount by summing the cyclic degradation amount and calendar degradation amount in the i-th cycle, and determines the dynamic coupling coefficient based on the coupling acceleration factor and the overall degradation amount. This allows the dynamic coupling coefficient to adaptively adjust as the battery aging degree changes. Compared to using a fixed coupling coefficient, this method can reflect the characteristic that the coupling strength between cyclic aging and calendar aging increases with the increase of the overall degradation amount, thus more accurately characterizing the aging coupling effect of the battery at different aging stages.

[0039] Step 240: Based on the product of the dynamic coupling coefficient of the i-th cycle, the cyclic decay of the i-th cycle, and the calendar decay of the i-th cycle, determine the cross-coupling term of the i-th cycle. The cross-coupling term is used to indicate the positive feedback coupling effect between cyclic aging and calendar aging that increases with the degree of aging.

[0040] In this embodiment, the cross-coupling term is used to characterize the interaction between cyclic aging and calendar aging. The positive feedback coupling effect here can be understood as follows: cyclic aging and calendar aging mutually promote each other, making the total aging effect of the battery greater than the sum of the two. For example, in the early stages of battery operation, both cyclic decay and calendar decay are small, and the cross-coupling term has a weak impact on the total decay; however, in the middle and later stages of battery operation, both cyclic decay and calendar decay have accumulated to a high level, at which point the cross-coupling term increases significantly, indicating that both aging mechanisms jointly accelerate the decline in energy efficiency. In actual operation, cyclic aging and calendar aging are not completely independent. For example, frequent deep cycling may lead to increased internal resistance, loss of active materials, or thickening of the interface film, making the battery more prone to calendar decay during subsequent resting or high SOC residence stages; conversely, prolonged exposure to high temperature or high SOC may alter the internal material state of the battery, making capacity decay and energy efficiency decline more pronounced during subsequent charge-discharge cycles. Therefore, there may be a mutually promoting relationship between cyclic aging and calendar aging. In this application, the aforementioned mutually promoting relationship is described through the cross-coupling term. The cross-coupling term can increase with the increase of cycle decay and calendar decay, and is used to characterize the enhanced coupling effect between cycle aging and calendar aging as the battery aging degree gradually deepens.

[0041] In some embodiments, the cross-coupling term can be used as an additional attenuation term when predicting energy efficiency, to correct prediction biases caused by a single aging factor or simply superimposed aging factors.

[0042] In some embodiments, the product of the dynamic coupling coefficient of the i-th cycle, the cyclic decay of the i-th cycle, and the calendar decay of the i-th cycle is determined as the cross-coupling term of the i-th cycle. For example, the cross-coupling term of the i-th cycle can be denoted as D_cross(i), the dynamic coupling coefficient of the i-th cycle as λ(i), the cyclic decay of the i-th cycle as D_cyc(i), and the calendar decay of the i-th cycle as D_cal(i). It is determined as follows: D_cross(i) = λ(i) × Dcyc_i × Dcal_i. As can be seen from the above formula, the cross-coupling term depends on both the cyclic decay and the calendar decay. When either the cyclic decay or the calendar decay is small, the cross-coupling term is small, indicating a weaker coupling effect between the two types of aging; when both the cyclic decay and the calendar decay increase, the cross-coupling term increases accordingly, indicating a stronger mutual promotion effect between cyclic aging and calendar aging.

[0043] Step 250: Based on the cyclic decay amount of the i-th period, the calendar decay amount of the i-th period, and the cross-coupling term of the i-th period, predict the energy efficiency corresponding to the i-th period.

[0044] Energy efficiency can be used to characterize the energy conversion efficiency of a battery energy storage system during the charging and discharging process in the i-th cycle, or it can be used as an efficiency index to characterize the degree of energy loss in a battery energy storage system. This application does not limit it in this respect.

[0045] In some embodiments, the overall efficiency decay factor for the i-th period is determined based on the cyclic decay, calendar decay, and cross-coupling term of the i-th period. The baseline energy efficiency is then corrected based on this overall efficiency decay factor to obtain the energy efficiency corresponding to the i-th period. For example, the cyclic decay for the i-th period is denoted as D_cyc(i), the calendar decay for the i-th period is denoted as D_cal(i), the dynamic coupling coefficient for the i-th period is denoted as λ(i), and the baseline energy efficiency for the i-th period is denoted as η0. The cross-coupling term for the i-th period can be expressed as D_cross(i) = λ(i) × D_cyc(i) × D_cal(i). Based on this, the overall efficiency decay factor corresponding to the i-th period can be expressed as D_eff(i) = D_cyc(i) + D_cal(i) + D_cross(i). That is: D_eff(i) = D_cyc(i) + D_cal(i) + λ(i)× D_cyc(i) × D_cal(i). Here, D_eff(i) characterizes the degree of energy efficiency degradation caused by cyclic aging, calendar aging, and their cross-coupling effects within the i-th cycle. Further, the energy efficiency corresponding to the i-th cycle can be expressed as: η(i) = η0× [1 - D_eff(i)]. Expanding D_eff(i), we get: η(i) = η0× [1 -D_cyc(i) - D_cal(i) - λ(i) × D_cyc(i) × D_cal(i)]. The above formula indicates that cyclic degradation, calendar degradation, and the cross-coupling term all reduce the energy efficiency of the battery storage system. Among them, cycle decay reflects the efficiency decay caused by charge-discharge cycles, calendar decay reflects the efficiency decay caused by factors such as time, temperature and SOC residence, and cross-coupling term reflects the additional efficiency decay caused by the mutual promotion between cycle aging and calendar aging.

[0046] In some embodiments, to avoid the predicted energy efficiency from exceeding a reasonable physical range, boundary constraints can be applied to the predicted η(i). For example, a lower limit η_min and an upper limit η_max of energy efficiency can be set, and the predicted energy efficiency can be limited to the range [η_min, η_max]: η(i) = min{η_max, max[η_min, η0× (1- D_eff(i))]}. Here, η_min can be the minimum allowable energy efficiency of the battery energy storage system, and η_max can be the maximum energy efficiency of the battery energy storage system under healthy conditions.

[0047] In some embodiments, if D_cyc(i), D_cal(i), or D_cross(i) are dimensionless decay ratios, they can be directly substituted into the above formulas; if the cyclic decay, calendar decay, or cross-coupling term is a capacity decay, internal resistance increase, or other dimensional index, it can be normalized first or converted into an efficiency decay ratio through a mapping function before being used to predict energy efficiency.

[0048] By employing the above method, this application, when predicting the energy efficiency corresponding to the i-th cycle, not only considers the independent effects of cyclic aging and calendar aging, but also further considers the cross-coupling effect between the two. By introducing a cross-coupling term, the positive feedback coupling effect between cyclic aging and calendar aging, which intensifies with the deepening of aging, can be reflected, thereby improving the accuracy of energy efficiency prediction results and further improving the accuracy of subsequent carbon emission predictions.

[0049] Step 260: Based on the energy efficiency corresponding to the i-th cycle, predict the carbon emissions corresponding to the i-th cycle.

[0050] In battery energy storage systems, decreased energy efficiency leads to greater energy loss when performing the same energy storage service. This energy loss typically needs to be compensated by the grid or other power sources, thus generating corresponding indirect carbon emissions. This application calculates the energy loss in the i-th cycle based on the predicted energy efficiency and predicts carbon emissions by combining this with the electricity carbon emission factor.

[0051] In some embodiments, predicting the carbon emissions corresponding to the i-th cycle based on the energy efficiency corresponding to the i-th cycle includes: determining the power loss in the i-th cycle based on the available capacity of the i-th cycle, the energy efficiency corresponding to the i-th cycle, and the number of cycles in the i-th cycle; determining the power carbon emission factor corresponding to the i-th cycle based on the power carbon emission factor corresponding to the baseline cycle, the growth rate of the green electricity penetration rate of the power grid, and the time interval between the i-th cycle and the baseline cycle, wherein the power carbon emission factor corresponding to the i-th cycle is negatively correlated with the time interval; and determining the carbon emissions corresponding to the i-th cycle based on the power loss in the i-th cycle and the power carbon emission factor corresponding to the i-th cycle.

[0052] The available capacity in the i-th cycle represents the effective capacity of the battery energy storage system capable of participating in charge-discharge operation within the i-th cycle. The available capacity can be determined based on the battery's rated capacity, capacity retention rate, SOH (State of Health), dischargeable capacity, or a capacity estimate reported by the BMS. The cycle number in the i-th cycle represents the number of charge-discharge cycles completed by the battery energy storage system within the i-th cycle; it can also be the equivalent cycle number or the weighted equivalent cycle number, which is not limited in this application.

[0053] In some embodiments, the power loss E_loss(i) in the i-th cycle is determined based on the available capacity of the i-th cycle, the energy efficiency corresponding to the i-th cycle, and the number of cycles in the i-th cycle, as shown below: E_loss(i) = Cap(i) × n(i) × [1 - η(i)]; Where Cap(i) refers to the available capacity in the i-th cycle, n(i) refers to the number of cycles in the i-th cycle, and η(i) refers to the energy efficiency corresponding to the i-th cycle. Optionally, if the unit of Cap(i) is kWh, n(i) is the number of cycles per day, 365 is the annualized number of days, and 365 / 1000 is the conversion of kgCO2 to tCO2 or kWh to MWh, then E_loss(i) = Cap(i) × n(i) × [1 - η(i)] × 365 / 1000.

[0054] In some embodiments, the electricity carbon emission factor corresponding to the i-th period is determined based on the electricity carbon emission factor corresponding to the baseline period, the growth rate of the grid green electricity penetration rate, and the time interval between the i-th period and the baseline period, as shown in the following formula: EF(i) = EF_base × (1 - r_green × t_i); Wherein, EF(i) refers to the electricity carbon emission factor corresponding to the i-th cycle, EF_base refers to the electricity carbon emission factor corresponding to the base cycle, r_green refers to the growth rate of the green electricity penetration rate of the power grid, and t_i refers to the time interval between the i-th cycle and the base cycle.

[0055] In some embodiments, the carbon emissions corresponding to the i-th cycle are determined based on the product of the power loss in the i-th cycle and the power carbon emission factor corresponding to the i-th cycle, as shown below: PCF(i) = E_loss(i)×EF(i), that is, PCF(i)= Cap(i) × n(i) × [1 - η(i)] × EF_base × (1 - r_green × t_i).

[0056] The method described above considers both the changes in the energy efficiency of the battery storage system itself and the changes in the external grid carbon emission factor when predicting carbon emissions for the i-th cycle. On one hand, determining the power loss based on available capacity, cycle count, and energy efficiency reflects the additional power loss caused by battery aging or system efficiency decline. On the other hand, determining the time-varying electricity carbon emission factor based on the grid's green electricity penetration rate reflects the decreasing trend of carbon emissions per unit of electricity during the grid's low-carbon transition. Therefore, compared to carbon emission prediction methods using fixed energy efficiency or fixed electricity carbon emission factors, this method improves the dynamic adaptability and accuracy of carbon emission prediction results.

[0057] In summary, the technical solution provided in this application does not directly predict the carbon emissions of a battery energy storage system using a fixed loss rate. Instead, it first determines the cycle decay and calendar decay for the i-th cycle based on the operating data within the i-th cycle, then determines the dynamic coupling coefficient based on the cycle decay and calendar decay, and further determines the cross-coupling term, thereby predicting the energy efficiency and carbon emissions corresponding to the i-th cycle. This avoids the inaccurate prediction problem caused by using a fixed loss rate to predict carbon emissions in related technologies, and improves the accuracy of energy efficiency and carbon emission prediction.

[0058] Furthermore, characterizing the coupling strength between cyclic aging and calendar aging using a dynamic coupling coefficient allows the coupling strength to increase with the degree of battery aging, avoiding the difficulty in adapting to different aging stages when using a fixed coupling coefficient, thereby improving the accuracy of battery aging state characterization. Simultaneously, characterizing the positive feedback coupling effect between cyclic aging and calendar aging, which increases with the degree of aging, using a cross-coupling term further reflects the mutually reinforcing effect between the two aging mechanisms. This allows for the introduction of the additional degradation effect resulting from the combined effect of cyclic aging and calendar aging in the energy efficiency prediction process, thereby further improving the accuracy of energy efficiency and carbon emission prediction results.

[0059] The following describes how to determine the cyclic decay amount for the i-th cycle based on the operating data within the i-th cycle.

[0060] In some embodiments, the operating data in the i-th cycle includes: the depth of discharge of each of the N types of discharge cycles in the i-th cycle, the average temperature of the i-th cycle, the charge / discharge rate of the i-th cycle, and the state of charge (SOC) fluctuation of the i-th cycle, where N is a positive integer.

[0061] The depth of discharge (DOD) characterizes the ratio of the amount of electricity released by the battery during a discharge cycle to its capacity. A higher DOD indicates greater cyclic stress on the battery during that discharge cycle.

[0062] In some embodiments, discharge cycles can be divided into N types of discharge cycles according to discharge depth, or into multiple discharge depth categories. Each type of discharge cycle can correspond to a discharge depth or a discharge depth range. For example, the discharge depth can be divided into multiple discharge depth ranges such as 0%–20%, 20%–40%, 40%–60%, 60%–80%, and 80%–100%, and the number of cycles within each discharge depth range can be counted. In this case, the k-th type of discharge cycle in the N types of discharge cycles can represent a group of discharge cycles with the same or similar discharge depths, which is not limited in this application.

[0063] Average temperature characterizes the thermal environment of the battery energy storage system or the average temperature of the battery itself during the i-th cycle. Charge / discharge rate characterizes the current stress of the battery during charging or discharging. State of charge (SOC) fluctuation characterizes the magnitude of the change in the battery's state of charge during the i-th cycle; a larger SOC fluctuation generally indicates a more drastic change in the battery's state of charge during charging or discharging.

[0064] In some embodiments, determining the cycle decay amount for the i-th cycle based on the operating data within the i-th cycle includes: determining the weighted equivalent cycle number for the i-th cycle based on the difference between the discharge depth of each of the N types of discharge cycles and the reference discharge depth within the i-th cycle; and determining the cycle decay amount for the i-th cycle based on the average temperature of the i-th cycle, the charge / discharge rate of the i-th cycle, the SOC fluctuation of the i-th cycle, and the weighted equivalent cycle number for the i-th cycle.

[0065] The reference depth of discharge is used as a benchmark value for calculating the contribution of cycle aging at different depths of discharge. For example, the reference depth of discharge can be taken as 80%, or it can be determined based on the battery type, cell chemistry, energy storage system application scenario, manufacturer calibration data, or historical aging test data.

[0066] The weighted equivalent cycle count is used to convert discharge cycles at different depths of discharge into equivalent cycle counts under a unified benchmark. Since different depths of discharge have varying effects on battery cyclic aging, directly using the cumulative cycle count may not distinguish the different contributions of deep and shallow discharge cycles to battery degradation. For example, a single high-depth discharge cycle may have a greater impact on battery cyclic aging than a single low-depth discharge cycle. Therefore, this application uses the difference between the discharge depth and a reference discharge depth to perform a weighted conversion for different discharge cycles.

[0067] In some embodiments, the cycle decay amount for the i-th cycle is determined based on the average temperature of the i-th cycle, the charge / discharge rate of the i-th cycle, the SOC fluctuation of the i-th cycle, and the weighted equivalent cycle number of the i-th cycle. It can be determined as follows: ; Where D_cyc(i) represents the cyclic decay amount in the i-th cycle; denoted by ; C_eff(i) represents the weighted equivalent number of cycles in the i-th cycle; β represents the decay exponent; Ea represents the apparent activation energy; R represents the gas constant; T_avg(i) represents the average temperature in the i-th cycle; CR(i) represents the charge / discharge rate in the i-th cycle; σ_SOC(i) represents the SOC fluctuation in the i-th cycle. δ represents the ratio stress factor; δ represents the SOC fluctuation stress factor.

[0068] In the above formula, Used to characterize the effect of the weighted equivalent number of cycles on the cycle decay; temperature correction term Used to characterize the effect of average temperature on cycle decay; operating stress correction term It is used to characterize the effect of charge / discharge rate and SOC fluctuation on cycle decay.

[0069] The above method determines the weighted equivalent number of cycles based on the difference between the discharge depth of each of the N types of discharge cycles in the i-th cycle and the reference discharge depth. This allows the discharge cycles at different discharge depths to be evaluated under a unified benchmark, avoiding the equal weighting of deep discharge cycles and shallow discharge cycles in the cycle decay, thus more accurately reflecting the differentiated contribution of different discharge depths to battery cycle aging.

[0070] Furthermore, by combining average temperature, charge / discharge rate, and SOC fluctuations to determine cycle degradation, the effects of thermal stress, current stress, and state-of-charge fluctuations on battery aging can be incorporated into the cycle degradation calculation process, making the obtained cycle degradation more consistent with the actual operating conditions of battery energy storage systems. Since cycle degradation also participates in subsequent calculations of dynamic coupling coefficients, cross-coupling terms, energy efficiency, and carbon emissions, this method can improve the accuracy of energy efficiency and carbon emission predictions.

[0071] In some embodiments, the operating data within the i-th cycle further includes: the number of cycles for each of the N types of discharge cycles within the i-th cycle. In some embodiments, determining the weighted equivalent cycle count for the i-th cycle based on the difference between the discharge depth of each of the N types of discharge cycles and the reference discharge depth within the i-th cycle includes: obtaining the discharge depth stress index; for the k-th type of discharge cycle among the N types of discharge cycles within the i-th cycle, determining the ratio between the discharge depth of the k-th type of discharge cycle and the reference discharge depth, where k is a positive integer less than or equal to N; determining the equivalent cycle corresponding to the k-th type of discharge cycle based on the ratio, the discharge depth stress index, and the number of cycles for the k-th type of discharge cycle; and obtaining the weighted equivalent cycle count for the i-th cycle based on the equivalent cycles corresponding to the N types of discharge cycles.

[0072] N types of discharge cycles can be multiple categories of discharge cycles divided according to discharge depth. Each type of discharge cycle corresponds to a discharge depth or a discharge depth range. The number of cycles in the k-th type of discharge cycle is used to indicate the number of discharge cycles belonging to the k-th type of discharge cycle in the i-th period. For example, the discharge cycles in the i-th period can be divided into multiple discharge depth ranges according to discharge depth, such as 0%–20%, 20%–40%, 40%–60%, 60%–80%, and 80%–100%. For each discharge depth range, the number of cycles in that range can be counted, and the representative discharge depth of that range can be used as the discharge depth of that type of discharge cycle. This representative discharge depth can be the median, average, or statistical value of the actual discharge depth in that range, and this application does not limit this.

[0073] For example, the discharge depth of the k-th type of discharge cycle within the i-th period is denoted as: The reference discharge depth is denoted as: Let the depth-of-discharge stress index be denoted as m, and the number of cycles for the k-th type of discharge cycle be denoted as: Then the equivalent cycle corresponding to the k-th type of discharge cycle can be expressed as: Where k is a positive integer less than or equal to N. The discharge depth stress index m is used to characterize the nonlinearity of the effect of discharge depth changes on cyclic aging. When m is greater than 1, it means that an increase in discharge depth will accelerate the increase in equivalent cyclic weight; when m equals 1, it means that the equivalent cyclic weight is approximately linearly related to the discharge depth; when m is less than 1, it means that the effect of discharge depth on equivalent cyclic weight is relatively mild.

[0074] In some embodiments, the weighted equivalent cycle number of the i-th cycle is determined based on the sum of the equivalent cycles corresponding to the N types of discharge cycles. Optionally, the sum of the equivalent cycles corresponding to the N types of discharge cycles is used as the weighted equivalent cycle number of the i-th cycle. As shown in the following formula: ; in, This represents the number of cycles for the k-th type of discharge cycle. In this way, discharge cycles at different depths can be uniformly converted into the equivalent number of cycles based on the reference discharge depth, making the calculation of cycle decay more consistent with actual operating conditions.

[0075] The above method, by introducing an equivalent cycle number weighted by discharge depth, avoids treating discharge cycles with different discharge depths equally, allowing the cycle degradation to more accurately characterize the actual cyclic aging stress experienced by the battery energy storage system in the i-th cycle. This weighted equivalent cycle number is further used to determine the cycle degradation, helping to improve the accuracy of predictions for dynamic coupling coefficients, cross-coupling terms, energy efficiency, and carbon emissions.

[0076] The following describes how to determine the calendar decay amount for the i-th period based on the running data within the i-th period.

[0077] In some embodiments, the operating data in the i-th cycle includes the operating time of the i-th cycle, the average temperature of the i-th cycle, and the SOC time series of the i-th cycle.

[0078] The operating time of the i-th cycle characterizes the duration for which the battery energy storage system is in operating, standby, quiescent, or online state during the i-th cycle. The average temperature characterizes the thermal environment or average temperature of the battery itself during the i-th cycle. The SOC time series characterizes the change in the battery's state of charge over time during the i-th cycle. The SOC time series can include multiple sampling times and their corresponding SOC values, for example: {(t1, SOC1), (t2, SOC2), ..., (tp, SOCp)}, where tp represents the p-th sampling time and SOCp represents the state of charge corresponding to the p-th sampling time.

[0079] Calendar degradation is primarily used to characterize the aging and degradation of a battery energy storage system over time due to factors such as temperature, state of charge (SOC), and operating time. Unlike cycle degradation, which is mainly affected by charge-discharge cycles, calendar degradation can occur even when the battery is not frequently charged and discharged. For example, prolonged exposure to high temperatures or high SOC ranges can lead to increased side reactions, interfacial film growth, loss of active lithium, or increased internal resistance, resulting in calendar degradation.

[0080] In some embodiments, determining the calendar decay amount for the i-th period based on the operating data within the i-th period includes: determining the cumulative residence time for at least one SOC interval based on the SOC time series of the i-th period, wherein the cumulative residence time of the SOC interval is used to indicate the residence duration of the SOC within the SOC interval; determining the SOC residence time distribution function for the i-th period based on the difference between the cumulative residence time of each of the at least one SOC interval and the operating time of the i-th period, wherein the SOC residence time distribution function is used to indicate the degree of influence of the residence duration distribution of the SOC in different SOC intervals within the i-th period on the calendar decay amount; and determining the calendar decay amount for the i-th period based on the operating time of the i-th period, the average temperature of the i-th period, and the SOC residence time distribution function of the i-th period.

[0081] The SOC range can be a predefined range of states of charge. For example, the SOC range can be divided into the following ranges: 0% to 20%, 20% to 40%, 40% to 60%, 60% to 80%, and 80% to 100%. It can also be divided into other ranges according to battery type, aging test data, operating strategy, or management needs. This application does not limit this.

[0082] For any SOC range, the cumulative residence time within that SOC range indicates the cumulative duration during which the battery's SOC is within that range in the i-th cycle. For example, if the battery is in the 80%–100% SOC range for 3 hours in the i-th cycle, then the corresponding cumulative residence time for that SOC range is 3 hours.

[0083] In one implementation, the cumulative dwell time of each SOC interval can be calculated based on the time interval between adjacent sampling points in the SOC time series. For example, for adjacent sampling times tp and tp+1, if the SOC value corresponding to time tp falls within a certain SOC interval, the time interval tp+1-tp can be added to the cumulative dwell time of that SOC interval. By iterating through all sampling points in the i-th period, the cumulative dwell time corresponding to each SOC interval can be obtained.

[0084] In some embodiments, the calendar decay amount for the i-th cycle is determined based on the runtime of the i-th cycle, the average temperature of the i-th cycle, and the SOC residence time distribution function of the i-th cycle. As shown in the following formula: ; in, This refers to the calendar aging factor, where t(i) represents the running time of the i-th cycle; Ea_cal represents the apparent activation energy corresponding to calendar aging; R represents the gas constant; and T_avg(i) represents the average temperature of the i-th cycle. ψ represents the temperature correction term, used to characterize the effect of average temperature on calendar aging rate; ψ represents the SOC dwell time influence coefficient, used to characterize the degree of enhancement effect of SOC dwell time distribution function on calendar decay; f_SOC_dwell(i) represents the SOC dwell time distribution function of the i-th period, used to characterize the degree of influence of SOC dwell time distribution in different SOC intervals on calendar decay within the i-th period.

[0085] The aforementioned method incorporates the residence duration distribution of State of Charge (SOC) across different SOC ranges into the calculation of calendar degradation, enabling calendar degradation to reflect the combined effects of operating time, average temperature, and SOC residence status. Compared to methods that predict calendar degradation solely based on operating time, average temperature, or average SOC, this approach more accurately characterizes the impact of prolonged residence at high SOC, prolonged residence at low SOC, or residence within specific SOC ranges on battery calendar aging, thus improving the accuracy of calendar degradation calculations. Furthermore, since calendar degradation is further involved in the prediction of dynamic coupling coefficients, cross-coupling terms, energy efficiency, and carbon emissions, this method can further improve the accuracy of energy efficiency and carbon emission predictions.

[0086] In some embodiments, determining the SOC dwell time distribution function for the i-th period based on the difference between the cumulative dwell time of each of at least one SOC interval and the running time of the i-th period includes: obtaining the weight coefficients of each of the at least one SOC interval; for any SOC interval among the at least one SOC intervals, determining the dwell time proportion corresponding to the SOC interval based on the ratio between the cumulative dwell time of the SOC interval and the running time of the i-th period; determining the weighted dwell value corresponding to the SOC interval based on the dwell time proportion corresponding to the SOC interval and the weight coefficient of the SOC interval; and summing the weighted dwell values ​​corresponding to the at least one SOC interval to obtain the SOC dwell time distribution function for the i-th period.

[0087] Different SOC ranges can correspond to different weighting coefficients, which characterize the intensity of the battery's impact on calendar aging when it is in the corresponding SOC range. For example, the weighting coefficient for the high SOC range can be greater than that for the medium SOC range, indicating that the battery is more prone to calendar aging when it is in a high SOC state for a long time. The weighting coefficients for low, medium, and high SOC ranges can be determined based on the battery chemistry system, manufacturer calibration data, historical operating data, experimental aging data, or maintenance experience.

[0088] In one implementation, the cumulative dwell time of the q-th SOC interval is denoted as t_q(i), the running time of the i-th cycle is denoted as t(i), and the weight coefficient of the q-th SOC interval is denoted as ω_q. Then, the dwell time proportion corresponding to the q-th SOC interval can be expressed as p_q(i) = t_q(i) / t(i). Further, the weighted dwell value corresponding to the SOC interval is determined based on the product of the dwell time proportion and the weight coefficient of the SOC interval. Optionally, the weighted dwell value corresponding to the q-th SOC interval can be expressed as V_q(i) = ω_q × p_q(i). By summing the weighted dwell values ​​corresponding to at least one SOC interval, the SOC dwell time distribution function for the i-th cycle can be obtained. For example, if the SOC dwell time distribution function for the i-th cycle is denoted as f_SOC_dwell(i), then the following formula is given: ; Where Q is the total number of at least one SOC interval, and q is a positive integer less than or equal to Q. As shown in the formula above, when the battery resides for a longer time in an SOC interval with a higher weighting coefficient, the proportion of residence time corresponding to that SOC interval is larger, and the weighted residence value is also larger, thus increasing f_SOC_dwell(i); when the battery mainly resides in an SOC interval with a lower weighting coefficient, f_SOC_dwell(i) is smaller. Therefore, the SOC residence time distribution function can be used to characterize the influence of the residence time distribution of different SOC intervals on the calendar decay in the i-th period.

[0089] In some embodiments, if the running time t(i) of the i-th cycle is zero or less than a preset time threshold, the SOC dwell time distribution function of that cycle may not be calculated, or the SOC dwell time distribution function of that cycle may be set to a preset value, or the SOC dwell time distribution function of an adjacent cycle may be used instead.

[0090] The method described above can determine the SOC residence time distribution function, which characterizes the influence of SOC residence status, based on the proportion of residence time in different SOC intervals and the corresponding weight coefficients of each SOC interval. Compared with methods that only use the average SOC or the SOC value at a certain moment, this method can more accurately distinguish the impact of short-term high SOC, long-term high SOC, and different combinations of SOC intervals on calendar aging, thereby improving the accuracy of calendar decay calculation.

[0091] The above embodiments introduced the method for using the multi-factor decay prediction model. The following embodiments will describe the training process of this multi-factor decay prediction model. The application and training of the multi-factor decay prediction model are related; details not elaborated in one embodiment can be found in the descriptions of the other embodiment.

[0092] Please refer to Figure 3 The diagram illustrates a flowchart of a model training method provided in one embodiment of this application. The execution entity for each step of this method can be a computer device; for example, the computer device can be... Figure 1 The model training device 110 in the computer system shown. The method may include at least one of the following steps (310-370).

[0093] Step 310: Obtain the operating data of the battery energy storage system in the i-th cycle, where i is a positive integer.

[0094] Step 320: Based on the running data within the i-th period, determine the cyclic decay amount and calendar decay amount for the i-th period using a multi-factor decay prediction model.

[0095] Step 330: Based on the cyclic decay amount and calendar decay amount of the i-th cycle, the dynamic coupling coefficient of the i-th cycle is determined by the multi-factor decay prediction model. The dynamic coupling coefficient is used to indicate the coupling strength between cyclic aging and calendar aging in the i-th cycle. The dynamic coupling coefficient is positively correlated with the cyclic decay amount and the calendar decay amount.

[0096] Step 340: Based on the product of the dynamic coupling coefficient of the i-th cycle, the cyclic decay amount of the i-th cycle, and the calendar decay amount of the i-th cycle, the cross-coupling term of the i-th cycle is determined by the multi-factor decay prediction model. The cross-coupling term is used to indicate the positive feedback coupling effect between cyclic aging and calendar aging that increases with the degree of aging.

[0097] Step 350: The energy efficiency corresponding to the i-th period is predicted by the multi-factor decay prediction model based on the cyclic decay amount of the i-th period, the calendar decay amount of the i-th period, and the cross-coupling term of the i-th period.

[0098] Step 360: Based on the energy efficiency corresponding to the i-th cycle, predict the carbon emissions corresponding to the i-th cycle.

[0099] Step 370: If the prediction deviation meets the preset conditions, adjust the correctable parameters in the multi-factor decay prediction model based on the prediction deviation to obtain the trained multi-factor decay prediction model; wherein, the prediction deviation is determined based on the deviation between the carbon emissions corresponding to the i-th period and the actual carbon emissions corresponding to the i-th period.

[0100] The actual carbon emissions corresponding to the i-th cycle can be determined based on the actual power loss of the battery energy storage system, the metered power, the actual power carbon emission factor, carbon metering system data, or manually calibrated data. This application does not limit this.

[0101] In some embodiments, if the prediction deviation does not meet the preset conditions, the correctable parameters in the multi-factor decay prediction model can be kept unchanged, or only a small, smooth update can be made to the correctable parameters. This avoids frequent adjustments to model parameters when the prediction error is small, reducing the risk of model oscillation.

[0102] In summary, the technical solution provided in this application does not directly predict the carbon emissions of a battery energy storage system using a fixed loss rate. Instead, it first determines the cycle decay and calendar decay for the i-th cycle based on the operating data within the i-th cycle, then determines the dynamic coupling coefficient based on the cycle decay and calendar decay, and further determines the cross-coupling term, thereby predicting the energy efficiency and carbon emissions corresponding to the i-th cycle. This avoids the inaccurate prediction problem caused by using a fixed loss rate to predict carbon emissions in related technologies, and improves the accuracy of energy efficiency and carbon emissions.

[0103] Furthermore, characterizing the coupling strength between cyclic aging and calendar aging using a dynamic coupling coefficient allows the coupling strength to increase with the degree of battery aging, avoiding the difficulty in adapting to different aging stages when using a fixed coupling coefficient, thereby improving the accuracy of battery aging state characterization. Simultaneously, characterizing the positive feedback coupling effect between cyclic aging and calendar aging, which increases with the degree of aging, using a cross-coupling term further reflects the mutually reinforcing effect between the two aging mechanisms. This allows for the introduction of the additional degradation effect resulting from the combined effect of cyclic aging and calendar aging in the energy efficiency prediction process, thereby further improving the accuracy of energy efficiency and carbon emission prediction results.

[0104] The following describes the specific implementation method of adjusting the correctable parameters in the multi-factor decay prediction model based on the prediction deviation, under the condition that the prediction deviation meets the preset conditions, to obtain the trained multi-factor decay prediction model.

[0105] In some embodiments, prediction bias includes systematic bias.

[0106] In some embodiments, when the prediction deviation meets a preset condition, adjusting the correctable parameters in the multi-factor decay prediction model based on the prediction deviation to obtain the trained multi-factor decay prediction model may include: determining the cycle deviation of the i-th cycle based on the difference between the carbon emissions corresponding to the i-th cycle and the actual carbon emissions corresponding to the i-th cycle; performing a moving average on the cycle deviations of each of the M consecutive cycles to obtain the system deviation, where M is an integer greater than 1, and the cycle deviations of each of the M consecutive cycles include the cycle deviation of the i-th cycle; and adjusting the correctable parameters based on the system deviation when the system deviation is greater than a first adaptive deviation threshold to obtain the trained multi-factor decay prediction model.

[0107] Systematic bias is used to characterize the persistent prediction deviation of a multi-factor decay prediction model over multiple consecutive periods. Compared with the prediction bias of a single period, systematic bias can mitigate the impact of random fluctuations, data acquisition errors, or abnormal operating conditions in a single period on model correction, making it more suitable for determining whether correctable parameters in the model need to be adjusted.

[0108] The periodic deviation is used to characterize the degree of deviation between the predicted carbon emissions and the actual carbon emissions corresponding to the i-th period. If the i-th period is an annual period, the deviation can be called the annual deviation; if the i-th period is a monthly period, a daily period, or other statistical period, the periodic deviation can also be called the monthly deviation, daily deviation, etc., and this application does not limit it in this way.

[0109] Optionally, the cycle deviation of the i-th cycle can be expressed as: ε(i) = PCF_pred(i) - PCF_actual(i). Wherein, PCF_pred(i) refers to the carbon emissions corresponding to the i-th cycle, and PCF_actual(i) refers to the actual carbon emissions corresponding to the i-th cycle.

[0110] Alternatively, to characterize only the magnitude of the deviation, the absolute deviation can also be used: ε(i) = |PCF_pred(i) - PCF_actual(i)|.

[0111] Alternatively, the relative deviation can be used as the periodic deviation, for example: ε(i)=|PCF_pred(i)-PCF_actual(i)| / PCF_actual(i).

[0112] After determining the periodic deviations of multiple consecutive periods, a moving average can be performed on the periodic deviations of each of the M consecutive periods to obtain the system deviation. In some embodiments, the system deviation is determined based on the average of the periodic deviations of each of the M consecutive periods. For example, if the system deviation is denoted as ε_sys, then the system deviation can be expressed as: ε_sys = [ε(i) + ε(i-1) + ... + ε(i-M+1)] / M. M can be configured according to the model update frequency, data stability, operating cycle length, or operation and maintenance strategy. For example, M=4 (taking the most recent 4 annual cycles), that is, taking the average of the periodic deviations of the most recent 4 cycles.

[0113] If the system deviation exceeds the first adaptive deviation threshold, the multi-factor decay prediction model can be considered to have persistent prediction deviation, requiring adaptive adjustment of the correctable parameters in the multi-factor decay prediction model. These correctable parameters may include the cyclic decay coefficient. Attenuation index β, apparent activation energy Ea, rate stress factor The correctable parameter may be at least one of the following: SOC fluctuation stress factor δ, SOC interval weighting coefficient, base coupling coefficient, coupling acceleration factor, and reference energy efficiency η0, but this application is not limited thereto. Optionally, the correctable parameter may include the cycle decay coefficient. The decay index β and the apparent activation energy Ea.

[0114] In one example, if the system bias characterizes the predicted carbon emissions as being lower than the actual carbon emissions over a long period, then at least one of the cycle decay coefficient, decay index, and apparent activation energy can be increased to improve the aging decay amount predicted in subsequent cycles, reduce the predicted energy efficiency, and thus increase the predicted carbon emissions.

[0115] In another example, if the system bias indicates that the predicted carbon emissions are higher than the actual carbon emissions in the long term, then at least one of the cycle decay coefficient, decay index and apparent activation energy can be reduced to reduce the aging decay amount predicted in subsequent cycles and improve the predicted energy efficiency, thereby reducing the predicted power loss and predicted carbon emissions.

[0116] In some embodiments, if the system deviation is less than or equal to a first adaptive deviation threshold, the correctable parameter is not adjusted.

[0117] The method described above determines the periodic bias by comparing the predicted and actual carbon emissions for the i-th period. It then performs a moving average on the periodic biases of M consecutive periods to obtain the systematic bias. This allows for the extraction of persistent bias directions from the prediction errors of multiple consecutive periods, mitigating the impact of single-period random fluctuations on model correction. When the systematic bias exceeds a first adaptive bias threshold, the correctable parameters in the multi-factor decay prediction model are adjusted based on the systematic bias, ensuring that parameter correction is only performed when persistent prediction deviations exist. This avoids frequent model parameter updates due to occasional anomalies in a single period, improves the stability of the model training process, and enhances the accuracy of the trained multi-factor decay prediction model in predicting carbon emissions for subsequent periods.

[0118] In some embodiments, the method further includes: determining a cumulative deviation based on the period deviations of at least two periods; and performing a moving average process on the period deviations of M consecutive periods to obtain a system deviation if the period deviation of the i-th period is greater than a first adaptive deviation threshold and the cumulative deviation is greater than a second adaptive threshold.

[0119] Periodic bias characterizes the degree of deviation between predicted and actual carbon emissions within a single period; cumulative bias characterizes the degree of deviation between the cumulative predicted and actual carbon emissions over multiple periods. By setting both periodic bias and cumulative bias simultaneously, single-period abnormal bias and long-term cumulative bias can be identified separately.

[0120] In some embodiments, the cumulative deviation can be determined based on the cumulative values ​​of predicted carbon emissions and actual carbon emissions over at least two periods. For example, the cumulative deviation... It can be represented as: ; Where s represents the period involved in the cumulative deviation calculation, and P is the number of at least two periods.

[0121] In some embodiments, if the period deviation ε(i) of the i-th period is greater than the first adaptive deviation threshold ε1, it indicates that there may be a large single-period prediction error in the current period, and further deviation analysis needs to be performed. If the cumulative deviation ε_cum(i) is greater than the second adaptive deviation threshold ε2, it indicates that the prediction error of multiple periods has reached a large level at the cumulative level, and even if the period deviation of a single period does not exceed the threshold, further deviation analysis needs to be performed. That is, the following triggering condition can be used: ε(i)>ε1 or ε_cum(i)>ε2. When at least one of the above conditions is met, the step of performing a moving average processing on the period deviations of each of the M consecutive periods to obtain the system deviation is performed.

[0122] The above method, by setting deviation analysis trigger conditions based on periodic and cumulative deviations, can simultaneously identify both single-period sudden deviations and multi-period cumulative deviations, avoiding the omission of long-term cumulative deviations when relying solely on single-period deviations. Furthermore, after the periodic or cumulative deviation meets the trigger conditions, a moving average is used to obtain the systematic deviation, and the need to adjust correctable parameters is determined based on the systematic deviation, distinguishing between systematic prediction deviations and random fluctuations. Thus, it can promptly detect prediction anomalies while avoiding frequent adjustments to model parameters due to occasional anomalies or noise in a single period, improving the stability of model calibration and the accuracy of carbon emission predictions.

[0123] In some embodiments, the method further includes: determining a random deviation based on the difference between the periodic deviation and the systematic deviation of the i-th period; determining the model uncertainty of the multi-factor decay prediction model based on the random deviation, wherein the model uncertainty is used to indicate the degree to which the prediction results of the multi-factor decay prediction model are affected by random fluctuations; determining a first adaptive deviation threshold based on the model uncertainty; and determining a second adaptive deviation threshold based on the first adaptive deviation threshold.

[0124] Random bias is used to characterize the residual after removing systematic bias from periodic bias. In other words, systematic bias reflects the tendency of the model to consistently overestimate or underestimate over multiple consecutive periods, while random bias reflects non-persistent bias caused by factors such as occasional operating conditions, measurement errors, data noise, and short-term operational fluctuations. Let ε(i) be the periodic bias of the i-th period, ε_sys(i) be the systematic bias of the i-th period, and ε_rnd(i) be the random bias of the i-th period. Then, random bias can be expressed as: ε_rnd(i) = ε(i) - ε_sys(i).

[0125] In some embodiments, the model uncertainty of a multi-factor decay prediction model can be determined based on random bias. For example, the model uncertainty can be denoted as σ_model. Optionally, the standard deviation of the random bias can be used to determine the model uncertainty. Optionally, the root mean square value of the random bias can also be used to determine the model uncertainty; this application does not limit this. The larger the model uncertainty σ_model, the more significantly the model prediction results are affected by random fluctuations; the smaller the model uncertainty, the more stable the model prediction results are and the weaker the random fluctuations.

[0126] In some embodiments, a first adaptive bias threshold can be determined based on model uncertainty. For example, if the first adaptive bias threshold is denoted as ε1, then: ε1 = k1 × σ_model. Where k1 is the first threshold adjustment coefficient. In this way, the first adaptive bias threshold can be automatically adjusted according to the degree of random fluctuation in the model. When the model's random fluctuation is large, σ_model is large, and the first adaptive bias threshold increases accordingly, thereby avoiding frequent triggering of bias analysis or parameter correction due to random noise; when the model's random fluctuation is small, σ_model is small, and the first adaptive bias threshold decreases accordingly, thereby improving the sensitivity to the actual prediction bias.

[0127] In some implementations, a second adaptive deviation threshold can be determined based on a first adaptive deviation threshold. For example, if the second adaptive deviation threshold is denoted as ε2, then: ε2 = k2 × ε1. Here, k2 is the adjustment coefficient for the second threshold. The second adaptive deviation threshold can be used to determine whether cumulative deviation triggers deviation analysis. Since cumulative deviation is used to characterize the long-term deviation between the predicted cumulative carbon emissions and the actual cumulative carbon emissions over multiple periods, its trigger threshold can be amplified or reduced using k2 to adapt to different model stability requirements or operational strategies.

[0128] The method described above decomposes periodic deviations into systematic and random deviations, determines model uncertainty based on random deviations, and then determines an adaptive deviation threshold based on the model uncertainty. This allows the model correction mechanism to distinguish between persistent prediction deviations and random fluctuations, avoiding frequent model parameter adjustments when random fluctuations are large, and promptly identifying systematic or cumulative deviations when random fluctuations are small. This improves the parameter correction stability of the multi-factor decay prediction model and enhances the accuracy of subsequent carbon emission prediction results.

[0129] like Figure 4 As shown, it illustrates a flowchart of a model training method provided in another embodiment of this application, which may include the following steps S1 to S6.

[0130] Step S1: Determine the annual deviation for the i-th cycle based on the difference between the carbon emissions corresponding to the i-th cycle and the actual carbon emissions corresponding to the i-th cycle.

[0131] Step S2: Determine the cumulative deviation based on the annual deviations of at least two periods.

[0132] Step S3: Determine whether the annual deviation is greater than the first adaptive deviation threshold and / or whether the cumulative deviation is greater than the second adaptive deviation threshold. If the conditions of step S3 are met, proceed to step S4 below; otherwise, terminate.

[0133] Step S4: Perform a moving average process on the annual deviations of each of the M consecutive periods to obtain the systematic deviation.

[0134] Step S5: Determine whether the system deviation is greater than the first adaptive deviation threshold. If the determination condition of step S5 is met, proceed to step S6 below; otherwise, terminate.

[0135] Step S6: Adjust the correctable parameters based on the system bias to obtain the trained multi-factor decay prediction model.

[0136] The above steps are illustrated by taking the period deviation of the i-th period as the annual deviation of the i-th period. This application does not limit the specific deviation type of the period deviation of the i-th period, that is, the annual deviation of the i-th period can be replaced by the period deviation of the i-th period.

[0137] For details not described in this embodiment, please refer to the preceding / following text; they will not be repeated here.

[0138] In some embodiments, adjusting the correctable parameters based on system bias to obtain a trained multi-factor degradation prediction model includes: determining the parameter adjustment amount of the correctable parameters based on system bias, the correction step size of the correctable parameters, and the sensitivity direction of the carbon emissions corresponding to the i-th cycle to the correctable parameters; wherein the correction step size of the correctable parameters is used to indicate the adjustment magnitude of the correctable parameters during the parameter adjustment process, and the sensitivity direction is used to determine the adjustment direction of the correctable parameters during the parameter adjustment process; adjusting the correctable parameters based on the parameter adjustment amount and the physical constraints of the correctable parameters to obtain a trained multi-factor degradation prediction model; wherein the physical constraints of the correctable parameters are used to indicate the allowable value range of the correctable parameters under the battery aging mechanism.

[0139] The calibration step size of a calibrable parameter indicates the magnitude of adjustment during parameter tuning. Different calibrable parameters can correspond to different calibration step sizes. For example, parameters that are more sensitive to carbon emission prediction results can be set with smaller calibration step sizes to avoid drastic changes in model parameters; parameters that have a more gradual impact on carbon emission prediction results can be set with relatively larger calibration step sizes to improve model convergence speed.

[0140] The sensitivity direction is used to determine the adjustment direction of the correctable parameter during the parameter adjustment process. The sensitivity direction can be determined based on the direction of the partial derivative of the carbon emissions corresponding to the i-th period with respect to the correctable parameter. For example, for the correctable parameter θ_j, the sensitivity direction of the predicted carbon emissions corresponding to the i-th period can be expressed as: sign( PCF(i) / θ_j). Where PCF(i) represents the predicted carbon emissions for the i-th cycle, θ_j represents the j-th correctable parameter, and sign(·) represents the sign function. PCF(i) / When θ_j) is positive, it means that increasing this correctable parameter will increase the predicted carbon emissions; when sign( PCF(i) / When θ_j) is negative, it means that increasing this correctable parameter will reduce the predicted carbon emissions.

[0141] In one implementation, the adjustment amount of the correctable parameter can be determined based on the following: Δθ_j = -μ_j × ε_sys(i) × sign( PCF(i) / θ_j) × 1 / (1+N_corr). Where, Δθ_j represents the parameter adjustment amount of the j-th correctable parameter; μ_j represents the correction step size corresponding to the j-th correctable parameter; ε_sys(i) represents the system deviation corresponding to the i-th period; sign( PCF(i) / θ_j) represents the direction of sensitivity of the predicted carbon emissions to the j-th correctable parameter; N_corr represents the number of parameter corrections performed.

[0142] In some embodiments, 1 / (1+N_corr) is used to decay the parameter adjustment amount. As the number of parameter calibrations increases, the amount of parameter adjustment per cycle gradually decreases, thereby reducing the risk of oscillations in the calibrable parameters in the later stages of training and improving the stability of the model calibration process.

[0143] After determining the parameter adjustment amount, the correctable parameter can be updated based on that adjustment amount. For example, the correctable parameter before the update can be denoted as (θ_j, old), and the correctable parameter after the update can be denoted as (θ_j, temp), then: (θ_j, temp) = (θ_j, old) + Δθ_j.

[0144] Temporary parameter values ​​can be constrained based on the physical constraints of the correctable parameters. These physical constraints indicate the allowable range of values ​​for the correctable parameters under the battery aging mechanism. By setting physical constraints, parameter values ​​that do not conform to the battery aging mechanism, physical meaning, or engineering experience can be avoided after parameter adjustment. For example, for the apparent activation energy Ea, its physical constraint can be expressed as: Ea ∈ [Ea_min, Ea_max]. For example, for the cycle decay calibration α, its physical constraint can be expressed as: α ∈ (0, α_max], etc., which are not limited in this application.

[0145] In some embodiments, if the updated temporary parameter value is within the corresponding physical constraint range, the temporary parameter value can be determined as the updated correctable parameter; if the temporary parameter value exceeds the corresponding physical constraint range, it can be truncated to the boundary value of the corresponding physical constraint range.

[0146] In some embodiments, different physical constraint methods may be adopted according to different types of calibrable parameters. For example, for parameters that must be positive, they can be constrained to be greater than 0; for parameters representing proportions or exponents, they can be constrained to a preset range; for parameters related to the battery material system, their allowable value range can be determined based on manufacturer calibration data or experimental data, and this application does not limit this.

[0147] The method described above, when adjusting the correctable parameters in the multi-factor decay prediction model, not only utilizes systematic bias to determine the model's persistent prediction deviation, but also combines the correction step size and sensitivity direction of the correctable parameters to determine the parameter adjustment amount, ensuring that parameter adjustments are directed towards reducing prediction bias. Simultaneously, by physically constraining the range of adjusted parameters, the physical rationality of the battery aging model can be avoided by preventing parameter updates from undermining its inherent logic. Therefore, the training stability, physical consistency, and subsequent carbon emission prediction accuracy of the multi-factor decay prediction model can be improved.

[0148] The following are embodiments of the apparatus described in this application, which can be used to execute the embodiments of the method described in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the method described in this application.

[0149] Please refer to Figure 5 This diagram illustrates a block diagram of a carbon emission prediction apparatus according to an embodiment of this application. The apparatus has the functionality to implement the aforementioned carbon emission prediction method example; this functionality can be implemented in hardware or by hardware executing corresponding software. The apparatus can be the computer device described above, or it can be installed within a computer device. Figure 5 As shown, the device 500 may include: an acquisition module 510, a first determination module 520, a second determination module 530, a third determination module 540, a first prediction module 550, and a second prediction module 560.

[0150] The acquisition module 510 is used to acquire the operating data of the battery energy storage system in the i-th cycle, where i is a positive integer.

[0151] The first determining module 520 is used to determine the cyclic decay amount of the i-th cycle based on the operating data within the i-th cycle, and to determine the calendar decay amount of the i-th cycle based on the operating data within the i-th cycle.

[0152] The second determining module 530 is used to determine the dynamic coupling coefficient of the i-th period based on the cyclic decay amount and the calendar decay amount of the i-th period. The dynamic coupling coefficient is used to indicate the coupling strength between cyclic aging and calendar aging in the i-th period. The dynamic coupling coefficient is positively correlated with the cyclic decay amount and the calendar decay amount.

[0153] The third determining module 540 is used to determine the cross-coupling term of the i-th period based on the product of the dynamic coupling coefficient of the i-th period, the cyclic decay amount of the i-th period, and the calendar decay amount of the i-th period. The cross-coupling term is used to indicate the positive feedback coupling effect between cyclic aging and calendar aging that increases with the degree of aging.

[0154] The first prediction module 550 is used to predict the energy efficiency corresponding to the i-th period based on the cyclic decay amount of the i-th period, the calendar decay amount of the i-th period, and the cross-coupling term of the i-th period.

[0155] The second prediction module 560 is used to predict the carbon emissions corresponding to the i-th cycle based on the energy efficiency corresponding to the i-th cycle.

[0156] In some embodiments, the second determining module 530 is used to determine a comprehensive decay amount based on the sum of the cyclic decay amount of the i-th period and the calendar decay amount of the i-th period; and to determine a dynamic coupling coefficient of the i-th period based on the product between the coupling acceleration factor and the comprehensive decay amount, wherein the coupling acceleration factor is used to indicate the degree of enhancement effect of the comprehensive decay amount on the dynamic coupling coefficient.

[0157] In some embodiments, the operating data in the i-th cycle includes: the discharge depth of each of the N types of discharge cycles in the i-th cycle, the average temperature of the i-th cycle, the charge / discharge rate of the i-th cycle, and the SOC fluctuation of the i-th cycle, where N is a positive integer; the first determining module 520 is used to determine the weighted equivalent cycle number of the i-th cycle based on the difference between the discharge depth of each of the N types of discharge cycles in the i-th cycle and the reference discharge depth; and to determine the cycle decay of the i-th cycle based on the average temperature of the i-th cycle, the charge / discharge rate of the i-th cycle, the SOC fluctuation of the i-th cycle, and the weighted equivalent cycle number of the i-th cycle.

[0158] In some embodiments, the first determining module 520 is used to obtain the discharge depth stress index; for the k-th type of discharge cycle in the N types of discharge cycles within the i-th period, determine the ratio between the discharge depth of the k-th type of discharge cycle and the reference discharge depth, where k is a positive integer less than or equal to N; based on the ratio, the discharge depth stress index, and the number of cycles of the k-th type of discharge cycle, determine the equivalent cycle corresponding to the k-th type of discharge cycle; and based on the equivalent cycles corresponding to the N types of discharge cycles, obtain the weighted equivalent cycle number of the i-th period.

[0159] In some embodiments, the operating data in the i-th period includes the operating time of the i-th period, the average temperature of the i-th period, and the SOC time series of the i-th period; the first determining module 520 is configured to determine the cumulative residence time of at least one SOC interval based on the SOC time series of the i-th period, wherein the cumulative residence time of the SOC interval is used to indicate the residence duration of the SOC in the SOC interval; determine the SOC residence time distribution function of the i-th period based on the difference between the cumulative residence time of the at least one SOC interval and the operating time of the i-th period, wherein the SOC residence time distribution function is used to indicate the degree of influence of the residence duration distribution of the SOC in different SOC intervals in the i-th period on the calendar decay amount; and determine the calendar decay amount of the i-th period based on the operating time of the i-th period, the average temperature of the i-th period, and the SOC residence time distribution function of the i-th period.

[0160] In some embodiments, the first determining module 520 is used to obtain the weight coefficients of each of the at least one SOC interval; for any SOC interval among the at least one SOC interval, based on the ratio between the cumulative dwell time of the SOC interval and the running time of the i-th cycle, determine the dwell time proportion corresponding to the SOC interval; based on the dwell time proportion corresponding to the SOC interval and the weight coefficient of the SOC interval, determine the weighted dwell value corresponding to the SOC interval; and accumulate the weighted dwell values ​​corresponding to the at least one SOC interval to obtain the SOC dwell time distribution function of the i-th cycle.

[0161] In some embodiments, the second prediction module 560 is configured to determine the power loss in the i-th cycle based on the available capacity of the i-th cycle, the energy efficiency corresponding to the i-th cycle, and the number of cycles in the i-th cycle; determine the power carbon emission factor corresponding to the i-th cycle based on the power carbon emission factor corresponding to the reference cycle, the growth rate of the green electricity penetration rate of the power grid, and the time interval between the i-th cycle and the reference cycle, wherein the power carbon emission factor corresponding to the i-th cycle is negatively correlated with the time interval; and determine the carbon emissions corresponding to the i-th cycle based on the power loss in the i-th cycle and the power carbon emission factor corresponding to the i-th cycle.

[0162] Please refer to Figure 6 This diagram illustrates a block diagram of a model training apparatus according to an embodiment of this application. The apparatus has the functionality to implement the model training method example described above; this functionality can be implemented in hardware or by hardware executing corresponding software. The apparatus can be the computer device described above, or it can be located within a computer device. Figure 6 As shown, the device 600 may include: an acquisition module 610, a first determination module 620, a second determination module 630, a third determination module 640, a first prediction module 650, a second prediction module 660, and a result module 670.

[0163] The acquisition module 610 is used to acquire the operating data of the battery energy storage system in the i-th cycle, where i is a positive integer.

[0164] The first determining module 620 is used to determine the cyclic decay amount and the calendar decay amount of the i-th period based on the running data in the i-th period using a multi-factor decay prediction model.

[0165] The second determining module 630 is used to determine the dynamic coupling coefficient of the i-th period based on the cyclic decay amount and the calendar decay amount of the i-th period through the multi-factor decay prediction model. The dynamic coupling coefficient is used to indicate the coupling strength between cyclic aging and calendar aging in the i-th period. The dynamic coupling coefficient is positively correlated with the cyclic decay amount and the calendar decay amount.

[0166] The third determining module 640 is used to determine the cross-coupling term of the i-th period based on the product of the dynamic coupling coefficient of the i-th period, the cyclic decay amount of the i-th period, and the calendar decay amount of the i-th period through the multi-factor decay prediction model. The cross-coupling term is used to indicate the positive feedback coupling effect between cyclic aging and calendar aging that increases with the degree of aging.

[0167] The first prediction module 650 is used to predict the energy efficiency corresponding to the i-th period based on the cyclic decay amount of the i-th period, the calendar decay amount of the i-th period, and the cross-coupling term of the i-th period through the multi-factor decay prediction model.

[0168] The second prediction module 660 is used to predict the carbon emissions corresponding to the i-th cycle based on the energy efficiency corresponding to the i-th cycle.

[0169] The obtaining module 670 is used to adjust the correctable parameters in the multi-factor decay prediction model based on the prediction deviation when the prediction deviation meets the preset conditions, so as to obtain the trained multi-factor decay prediction model; wherein the prediction deviation is determined based on the deviation between the carbon emissions corresponding to the i-th period and the actual carbon emissions corresponding to the i-th period.

[0170] In some embodiments, the prediction bias includes a systematic bias; the obtaining module 670 is used to determine the cycle bias of the i-th cycle based on the difference between the carbon emissions corresponding to the i-th cycle and the actual carbon emissions corresponding to the i-th cycle; perform a moving average processing on the cycle biases of each of the M consecutive cycles to obtain the systematic bias, where M is an integer greater than 1, and the cycle biases of each of the M consecutive cycles include the cycle bias of the i-th cycle; if the systematic bias is greater than a first adaptive bias threshold, adjust the correctable parameter based on the systematic bias to obtain the trained multi-factor decay prediction model.

[0171] In some embodiments, the obtaining module 670 is used to determine the parameter adjustment amount of the correctable parameter based on the system deviation, the correction step size of the correctable parameter, and the sensitivity direction of the carbon emissions corresponding to the i-th cycle to the correctable parameter; wherein, the correction step size of the correctable parameter is used to indicate the adjustment magnitude of the correctable parameter during the parameter adjustment process, and the sensitivity direction is used to determine the adjustment direction of the correctable parameter during the parameter adjustment process; based on the parameter adjustment amount of the correctable parameter and the physical constraints of the correctable parameter, the correctable parameter is adjusted to obtain the trained multi-factor decay prediction model; wherein, the physical constraints of the correctable parameter are used to indicate the allowable value range of the correctable parameter under the battery aging mechanism.

[0172] In some embodiments, the device 600 further includes: a fourth determining module ( Figure 6 (Not shown in the image).

[0173] The fourth determining module is used to determine the cumulative deviation based on the period deviation of at least two periods; and to perform the step of performing a moving average processing on the period deviation of each of the M consecutive periods to obtain the system deviation if the period deviation of the i-th period is greater than the first adaptive deviation threshold and the cumulative deviation is greater than the second adaptive threshold.

[0174] In some embodiments, the device 600 further includes: a fifth determining module ( Figure 6 (Not shown in the image).

[0175] The fifth determining module is used to determine a random deviation based on the difference between the period deviation of the i-th period and the system deviation; to determine the model uncertainty of the multi-factor decay prediction model based on the random deviation, wherein the model uncertainty is used to indicate the degree to which the prediction result of the multi-factor decay prediction model is affected by random fluctuations; to determine the first adaptive deviation threshold based on the model uncertainty; and to determine the second adaptive deviation threshold based on the first adaptive deviation threshold.

[0176] It should be noted that the apparatus provided in the above embodiments is only illustrated by the division of the above functional modules when implementing its functions. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the content structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0177] Please refer to Figure 7 The diagram shows a structural block diagram of a computer device 700 provided in one embodiment of this application.

[0178] Typically, computer device 700 includes a processor 710 and a memory 720.

[0179] Processor 710 may include one or more processing cores, such as a quad-core processor or an octa-core processor. Processor 710 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field Programmable Gate Array), and PLA (Programmable Logic Array). Processor 710 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, processor 710 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, processor 710 may also include an AI processor, which is used to handle computational operations related to machine learning.

[0180] The memory 720 may include one or more computer-readable storage media, which may be non-transitory. The memory 720 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory 720 are used to store a computer program configured to be executed by one or more processors to implement the carbon emission prediction method or the model training method described above.

[0181] Those skilled in the art will understand that Figure 7 The structure shown does not constitute a limitation on the computer device 700, and may include more or fewer components than shown, or combine certain components, or use different component arrangements.

[0182] In an exemplary embodiment, a computer-readable storage medium is also provided, wherein a computer program is stored in the storage medium, and the computer program is executed by a processor to implement the above-described carbon emission prediction method or the above-described model training method. Optionally, the computer-readable storage medium may include: ROM (Read-Only Memory), RAM (Random Access Memory), SSD (Solid State Drives), or optical disc, etc. The random access memory may include ReRAM (Resistance Random Access Memory) and DRAM (Dynamic Random Access Memory).

[0183] In an exemplary embodiment, a computer program product is also provided, the computer program product comprising a computer program executed by a processor to implement the above-described carbon emission prediction method or the above-described model training method.

[0184] It should be noted that the collection and processing of relevant data in this application (such as operational data within the i-th period) should strictly comply with the requirements of relevant national laws and regulations in actual application, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing within the scope of laws and regulations and the authorization of the personal information subject.

[0185] It should be understood that "multiple" as used herein refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. Furthermore, the step numbers described herein are merely illustrative of one possible execution order. In some other embodiments, the steps may not be executed in numerical order, such as two steps with different numbers being executed simultaneously, or two steps with different numbers being executed in the reverse order of the illustration. This application does not limit this.

[0186] The above description is merely an exemplary embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for predicting carbon emissions, characterized in that, The method includes: Obtain the operating data of the battery energy storage system in the i-th cycle, where i is a positive integer; Based on the operating data within the i-th cycle, determine the cyclic decay amount for the i-th cycle, and based on the operating data within the i-th cycle, determine the calendar decay amount for the i-th cycle. Based on the cyclic decay amount and the calendar decay amount of the i-th cycle, the dynamic coupling coefficient of the i-th cycle is determined. The dynamic coupling coefficient is used to indicate the coupling strength between cyclic aging and calendar aging within the i-th cycle. The dynamic coupling coefficient is positively correlated with the cyclic decay amount and the calendar decay amount. Based on the product of the dynamic coupling coefficient of the i-th cycle, the cyclic decay of the i-th cycle, and the calendar decay of the i-th cycle, the cross-coupling term of the i-th cycle is determined. The cross-coupling term is used to indicate the positive feedback coupling effect between cyclic aging and calendar aging that increases with the degree of aging. Based on the cyclic decay amount of the i-th cycle, the calendar decay amount of the i-th cycle, and the cross-coupling term of the i-th cycle, the energy efficiency corresponding to the i-th cycle is predicted; wherein, the energy efficiency is used to characterize the energy conversion efficiency and / or energy loss degree of the battery energy storage system during the charging and discharging process in the i-th cycle. Based on the available capacity of the i-th cycle, the energy efficiency corresponding to the i-th cycle, and the number of cycles of the i-th cycle, the power loss of the i-th cycle is determined; Based on the electricity carbon emission factor corresponding to the baseline period, the growth rate of green electricity penetration rate of the power grid, and the time interval between the i-th period and the baseline period, the electricity carbon emission factor corresponding to the i-th period is determined, and the electricity carbon emission factor corresponding to the i-th period is negatively correlated with the time interval. Based on the power loss in the i-th cycle and the corresponding carbon emission factor for the i-th cycle, the carbon emission amount for the i-th cycle is determined.

2. The method according to claim 1, characterized in that, The determination of the dynamic coupling coefficient for the i-th period based on the cyclic decay amount and the calendar decay amount for the i-th period includes: The overall attenuation is determined based on the sum of the cyclic attenuation amount of the i-th period and the calendar attenuation amount of the i-th period. The dynamic coupling coefficient of the i-th cycle is determined based on the product between the coupling acceleration factor and the overall attenuation, wherein the coupling acceleration factor is used to indicate the degree of enhancement effect of the overall attenuation on the dynamic coupling coefficient.

3. The method according to claim 1, characterized in that, The operating data in the i-th cycle includes: the discharge depth of each of the N types of discharge cycles in the i-th cycle, the average temperature of the i-th cycle, the charge / discharge rate of the i-th cycle, and the state of charge (SOC) fluctuation of the i-th cycle, where N is a positive integer; Determining the cyclic decay amount for the i-th cycle based on the operating data within the i-th cycle includes: Based on the difference between the discharge depth of each of the N types of discharge cycles in the i-th cycle and the reference discharge depth, the weighted equivalent cycle number of the i-th cycle is determined. The cycle decay amount of the i-th cycle is determined based on the average temperature of the i-th cycle, the charge / discharge rate of the i-th cycle, the SOC fluctuation of the i-th cycle, and the weighted equivalent cycle number of the i-th cycle.

4. The method according to claim 3, characterized in that, The determination of the weighted equivalent cycle number for the i-th cycle based on the difference between the discharge depth of each of the N types of discharge cycles within the i-th cycle and the reference discharge depth includes: Obtain the discharge depth stress index; For the kth type of discharge cycle in the N types of discharge cycles within the i-th period, determine the ratio between the discharge depth of the kth type of discharge cycle and the reference discharge depth, where k is a positive integer less than or equal to N; Based on the ratio, the discharge depth stress index, and the number of cycles of the k-th type of discharge cycle, the equivalent cycle corresponding to the k-th type of discharge cycle is determined; Based on the equivalent cycles corresponding to the N types of discharge cycles, the weighted equivalent cycle number of the i-th cycle is obtained.

5. The method according to claim 1, characterized in that, The operating data in the i-th cycle includes the operating time of the i-th cycle, the average temperature of the i-th cycle, and the SOC time series of the i-th cycle; Determining the calendar decay amount for the i-th period based on the operating data within the i-th period includes: Based on the SOC time series of the i-th period, the cumulative dwell time of at least one SOC interval is determined, and the cumulative dwell time of the SOC interval is used to indicate the dwell time of the SOC within the SOC interval. Based on the difference between the cumulative dwell time of each of the at least one SOC interval and the running time of the i-th cycle, the SOC dwell time distribution function of the i-th cycle is determined. The SOC dwell time distribution function is used to indicate the degree of influence of the dwell time distribution of SOC in different SOC intervals in the i-th cycle on the calendar decay amount. Based on the running time of the i-th cycle, the average temperature of the i-th cycle, and the SOC residence time distribution function of the i-th cycle, the calendar decay amount of the i-th cycle is determined.

6. The method according to claim 5, characterized in that, The step of determining the SOC dwell time distribution function for the i-th period based on the difference between the cumulative dwell time of each of the at least one SOC interval and the running time of the i-th period includes: Obtain the weight coefficients for each of the at least one SOC interval; For any SOC interval among the at least one SOC intervals, the proportion of the residence time corresponding to the SOC interval is determined based on the ratio between the cumulative residence time of the SOC interval and the running time of the i-th cycle. Based on the proportion of dwell time corresponding to the SOC interval and the weight coefficient of the SOC interval, the weighted dwell value corresponding to the SOC interval is determined; The weighted residence values ​​corresponding to the at least one SOC interval are accumulated to obtain the SOC residence time distribution function of the i-th period.

7. A model training method, characterized in that, The method includes: Obtain the operating data of the battery energy storage system in the i-th cycle, where i is a positive integer; Based on the running data within the i-th period, the cyclic decay amount and calendar decay amount of the i-th period are determined using a multi-factor decay prediction model. The multi-factor decay prediction model determines the dynamic coupling coefficient of the i-th period based on the cyclic decay amount and the calendar decay amount of the i-th period. The dynamic coupling coefficient is used to indicate the coupling strength between cyclic aging and calendar aging within the i-th period. The dynamic coupling coefficient is positively correlated with the cyclic decay amount and the calendar decay amount. The multi-factor decay prediction model determines the cross-coupling term for the i-th period based on the product of the dynamic coupling coefficient, the cyclic decay amount, and the calendar decay amount for the i-th period. The cross-coupling term is used to indicate the positive feedback coupling effect between cyclic aging and calendar aging that increases with the degree of aging. The multi-factor decay prediction model predicts the energy efficiency corresponding to the i-th cycle based on the cyclic decay amount, the calendar decay amount, and the cross-coupling term of the i-th cycle; wherein, the energy efficiency is used to characterize the energy conversion efficiency and / or energy loss degree of the battery energy storage system during the charging and discharging process in the i-th cycle. Based on the available capacity of the i-th cycle, the energy efficiency corresponding to the i-th cycle, and the number of cycles of the i-th cycle, the power loss of the i-th cycle is determined; Based on the electricity carbon emission factor corresponding to the baseline period, the growth rate of green electricity penetration rate of the power grid, and the time interval between the i-th period and the baseline period, the electricity carbon emission factor corresponding to the i-th period is determined, and the electricity carbon emission factor corresponding to the i-th period is negatively correlated with the time interval. Based on the power loss in the i-th cycle and the power carbon emission factor corresponding to the i-th cycle, the carbon emission amount corresponding to the i-th cycle is determined; If the prediction deviation meets the preset conditions, the correctable parameters in the multi-factor decay prediction model are adjusted based on the prediction deviation to obtain the trained multi-factor decay prediction model; wherein, the prediction deviation is determined based on the deviation between the carbon emissions corresponding to the i-th period and the actual carbon emissions corresponding to the i-th period.

8. The method according to claim 7, characterized in that, The prediction bias includes systematic bias; The step of adjusting the correctable parameters in the multi-factor decay prediction model based on the prediction deviation, under the condition that the prediction deviation meets the preset conditions, to obtain the trained multi-factor decay prediction model, includes: The cycle deviation of the i-th cycle is determined based on the difference between the carbon emissions corresponding to the i-th cycle and the actual carbon emissions corresponding to the i-th cycle. The system deviation is obtained by performing a moving average on the period deviations of each of the M consecutive periods, where M is an integer greater than 1, and the period deviations of each of the M consecutive periods include the period deviation of the i-th period. If the system deviation is greater than the first adaptive deviation threshold, the correctable parameter is adjusted based on the system deviation to obtain the trained multi-factor decay prediction model.

9. The method according to claim 8, characterized in that, The step of adjusting the correctable parameters based on the system bias to obtain the trained multi-factor decay prediction model includes: Based on the system deviation, the correction step size of the correctable parameter, and the sensitivity direction of the carbon emissions corresponding to the i-th cycle to the correctable parameter, the parameter adjustment amount of the correctable parameter is determined; wherein, the correction step size of the correctable parameter is used to indicate the adjustment magnitude of the correctable parameter during the parameter adjustment process, and the sensitivity direction is used to determine the adjustment direction of the correctable parameter during the parameter adjustment process. Based on the parameter adjustment amount of the correctable parameter and the physical constraints of the correctable parameter, the correctable parameter is adjusted to obtain the trained multi-factor decay prediction model; wherein, the physical constraints of the correctable parameter are used to indicate the allowable value range of the correctable parameter under the battery aging mechanism.

10. The method according to claim 8, characterized in that, The method further includes: The cumulative deviation is determined based on the period deviation of at least two periods. If the period deviation of the i-th period is greater than at least one of the first adaptive deviation threshold and the cumulative deviation is greater than the second adaptive deviation threshold, the step of performing a moving average process on the period deviations of each of the M consecutive periods to obtain the system deviation is performed.

11. The method according to claim 10, characterized in that, The method further includes: The random deviation is determined based on the difference between the period deviation of the i-th period and the system deviation; Based on the random bias, the model uncertainty of the multi-factor decay prediction model is determined, and the model uncertainty is used to indicate the degree to which the prediction results of the multi-factor decay prediction model are affected by random fluctuations. Based on the model uncertainty, the first adaptive deviation threshold is determined; The second adaptive deviation threshold is determined based on the first adaptive deviation threshold.

12. A carbon emission prediction device, characterized in that, The device includes: The acquisition module is used to acquire the operating data of the battery energy storage system in the i-th cycle, where i is a positive integer; The first determining module is used to determine the cyclic decay amount of the i-th period based on the running data within the i-th period, and to determine the calendar decay amount of the i-th period based on the running data within the i-th period. The second determining module is used to determine the dynamic coupling coefficient of the i-th period based on the cyclic decay amount and the calendar decay amount of the i-th period. The dynamic coupling coefficient is used to indicate the coupling strength between cyclic aging and calendar aging in the i-th period. The dynamic coupling coefficient is positively correlated with the cyclic decay amount and the calendar decay amount. The third determining module is used to determine the cross-coupling term of the i-th period based on the product of the dynamic coupling coefficient of the i-th period, the cyclic decay amount of the i-th period, and the calendar decay amount of the i-th period. The cross-coupling term is used to indicate the positive feedback coupling effect between cyclic aging and calendar aging that increases with the degree of aging. The first prediction module is used to predict the energy efficiency corresponding to the i-th cycle based on the cyclic decay amount of the i-th cycle, the calendar decay amount of the i-th cycle, and the cross-coupling term of the i-th cycle; wherein, the energy efficiency is used to characterize the energy conversion efficiency and / or energy loss degree of the battery energy storage system during the charging and discharging process in the i-th cycle. The second prediction module is used to determine the power loss in the i-th cycle based on the available capacity in the i-th cycle, the energy efficiency corresponding to the i-th cycle, and the number of cycles in the i-th cycle. The second prediction module is further configured to determine the electricity carbon emission factor corresponding to the i-th period based on the electricity carbon emission factor corresponding to the benchmark period, the growth rate of the green electricity penetration rate of the power grid, and the time interval between the i-th period and the benchmark period, wherein the electricity carbon emission factor corresponding to the i-th period is negatively correlated with the time interval. The second prediction module is further configured to determine the carbon emissions corresponding to the i-th cycle based on the power loss in the i-th cycle and the power carbon emission factor corresponding to the i-th cycle.

13. A model training device, characterized in that, The device includes: The acquisition module is used to acquire the operating data of the battery energy storage system in the i-th cycle, where i is a positive integer; The first determining module is used to determine the cyclic decay amount and the calendar decay amount of the i-th period based on the running data in the i-th period using a multi-factor decay prediction model. The second determining module is used to determine the dynamic coupling coefficient of the i-th period based on the cyclic decay amount and the calendar decay amount of the i-th period through the multi-factor decay prediction model. The dynamic coupling coefficient is used to indicate the coupling strength between cyclic aging and calendar aging in the i-th period. The dynamic coupling coefficient is positively correlated with the cyclic decay amount and the calendar decay amount. The third determining module is used to determine the cross-coupling term of the i-th period based on the product of the dynamic coupling coefficient of the i-th period, the cyclic decay amount of the i-th period, and the calendar decay amount of the i-th period through the multi-factor decay prediction model. The cross-coupling term is used to indicate the positive feedback coupling effect between cyclic aging and calendar aging that increases with the degree of aging. The first prediction module is used to predict the energy efficiency corresponding to the i-th cycle based on the cyclic decay amount of the i-th cycle, the calendar decay amount of the i-th cycle, and the cross-coupling term of the i-th cycle through the multi-factor decay prediction model; wherein, the energy efficiency is used to characterize the energy conversion efficiency and / or energy loss degree of the battery energy storage system during the charging and discharging process in the i-th cycle. The second prediction module is used to determine the power loss in the i-th cycle based on the available capacity in the i-th cycle, the energy efficiency corresponding to the i-th cycle, and the number of cycles in the i-th cycle. The second prediction module is further configured to determine the electricity carbon emission factor corresponding to the i-th period based on the electricity carbon emission factor corresponding to the benchmark period, the growth rate of the green electricity penetration rate of the power grid, and the time interval between the i-th period and the benchmark period, wherein the electricity carbon emission factor corresponding to the i-th period is negatively correlated with the time interval. The second prediction module is further configured to determine the carbon emissions corresponding to the i-th cycle based on the power loss in the i-th cycle and the power carbon emission factor corresponding to the i-th cycle. The module is used to adjust the correctable parameters in the multi-factor decay prediction model based on the prediction deviation when the prediction deviation meets the preset conditions, so as to obtain the trained multi-factor decay prediction model; wherein the prediction deviation is determined based on the deviation between the carbon emissions corresponding to the i-th period and the actual carbon emissions corresponding to the i-th period.

14. A computer device, characterized in that, The computer device includes a processor and a memory, the memory storing a computer program that is loaded and executed by the processor to implement the method as claimed in any one of claims 1 to 6, or to implement the method as claimed in any one of claims 7 to 11.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that is loaded and executed by a processor to implement the method as claimed in any one of claims 1 to 6, or to implement the method as claimed in any one of claims 7 to 11.

16. A computer program product, characterized in that, The computer program product includes computer instructions that are executed by a processor to implement the method as claimed in any one of claims 1 to 6, or to implement the method as claimed in any one of claims 7 to 11.

Images