Method and device for monitoring state of storage battery of transformer substation

Abstract

The invention relates to the technical field of electrical variable measurement, in particular to a state monitoring method and device for a transformer substation storage battery, and the method comprises the steps: obtaining a first electrical variable curve of each storage battery in a future time period T1 through a first model; according to the health coefficients of the standby group and the power supply group, obtaining the switching suitability F1 of the standby group in the switching time; using a second model to obtain a second electrical transformation curve of each battery in a future time period T2; the first electrical transformation curve is corrected according to the second electrical transformation curve, and the standby group switching time t1 and the switching suitability F2 are obtained again through the corrected first electrical transformation curve; and updating the first electrical transformation curve according to the first difference between the F1 and the F2 and the first duration from the current moment to the time t1, and switching the power supply group into the standby group at the standby group switching time t1. The storage battery pack is switched by obtaining the future storage battery state, so that the storage battery can stably supply power to the power grid for a long time.

Description

Technical Field

[0001] The present invention relates to the technical field of electric variable measurement, and in particular to a method and device for monitoring the status of batteries in substations. Background Art

[0002] Substation battery packs play a vital role in situations or scenarios such as station-wide power outages, power equipment failures within the substation, and load shaving. Battery status monitoring (for example, monitoring changes in internal resistance) is a key link in maintaining stable grid operation. A key point is that when a substation uses battery packs for power supply, battery status monitoring can be used to predict the battery's status at future times. This advance knowledge of the future status can be used to determine the battery's subsequent health. If the battery's health is not as good as expected (or its status changes abnormally), the battery pack can be replaced in a timely manner to prevent batteries with abnormal status changes from affecting the power supply of the entire battery pack, thereby ensuring the long-term stable operation of the grid.

[0003] When predicting the battery status over a longer time scale in the future, the battery's health or abnormal state changes can be monitored more reliably. However, the prediction of battery status over a longer time scale often has large offset errors and requires a large amount of calculation and time (especially when the number of batteries in the battery pack is large), making it difficult to predict the status in real time. This can lead to untimely switching of battery packs and affect the long-term stable power supply of the battery to the grid.

[0004] In summary, when switching battery groups by predicting the battery status over a longer time scale in the future, there is a possibility that the battery group switching is not timely, affecting the long-term stable power supply of the battery to the power grid. Summary of the Invention

[0005] To solve the above problems, the present invention provides a method and device for monitoring the status of a substation battery.

[0006] The method and device for monitoring the status of substation batteries of the present invention adopt the following technical solutions: One embodiment of the present invention provides a method for monitoring the status of a battery in a substation, the method comprising the following steps: All batteries are divided into a power supply group and several backup groups; a first power change curve of each battery in a future time period T1 is obtained; the first power change curves of all batteries in the power supply group are used to evaluate the health factor of the power supply group at each moment, and a backup group switching time is obtained. At the backup group switching time, the difference between the health factor of all backup groups and the health factor of the power supply group is the largest, and the maximum value is recorded as the switching suitability F1; At the current moment in time period T1, a second electrical variation curve of each battery in a future time period T2 is obtained; based on the internal resistance of each battery in the power supply group that has been measured and the difference between the second electrical variation curve and the first electrical variation curve, the first electrical variation curve is corrected, and the standby group switching time t1 and the switching suitability F2 are re-obtained using the corrected first electrical variation curve; the duration of T1 is greater than T2; based on the first difference between F1 and F2 and the first duration from the current moment to time t1, the first electrical variation curve of each battery is updated, and at the same time, the power supply group is switched to the standby group at the standby group switching time t1.

[0007] Preferably, the specific steps of calculating the health coefficient of the power supply group at each moment are as follows: For the first electrical curve of any battery in the power supply group, read the internal resistance at time i in the first electrical curve , obtain the internal resistance growth rate at time i; the health index of the first electric curve at time i and the internal resistance growth rate at time i, were negatively correlated; Whenever the first electrical variation curve of any battery in the power supply group is obtained, for each moment on the first electrical variation curve, and for all the first electrical variation curves obtained in the power supply group, the minimum value of the health indicators of all the first electrical variation curves at each moment is used as the health coefficient of the power supply group at each moment.

[0008] Preferably, the step of obtaining the standby group switching time, wherein the difference between the health coefficients of all standby groups and the health coefficient of the power supply group is the largest under the standby group switching time, and the maximum value is recorded as the switching suitability F1, includes the following specific steps: Whenever the first power curve of each battery in the power supply group is obtained, for the time period T1 in which the first power curve is located, and for any moment in time period T1, the difference between the health coefficient of any standby group and the health coefficient of the power supply group at that moment is obtained, and recorded as the health difference of any standby group at that moment; Perform K-means clustering on the health differences of all backup groups at each moment and cluster them into two categories. The mean of all health differences in each category is recorded as the cluster center, and the category with the largest cluster center is recorded as the target category. Obtain the distribution consistency of the backup groups in the target category obtained at any moment and the adjacent moments; record the product of the cluster center of the target category at any moment and the distribution consistency as the second index at each moment; within the time period T1, the moment with the maximum second index is recorded as the backup group switching time, recorded as t, and the backup group with the largest health difference at the backup group switching time t is used as the target group; the maximum health difference is recorded as the switching suitability, represented by F1.

[0009] Preferably, the method of correcting the first electric change curve according to the internal resistance of each battery in the power supply group and the difference between the second electric change curve and the first electric change curve comprises the following specific steps: Any moment in the time period T1 is recorded as the target moment, and the internal resistance collected at all moments before the target moment in the time period T1 constitutes the actual internal resistance sequence; Any internal resistance in the actual internal resistance sequence is recorded as , x1 represents the acquisition time of the internal resistance, and the internal resistance at the time x1 is obtained on the first electrical curve, which is recorded as ,Will Recorded as the first prediction error; the average of the first prediction errors corresponding to all internal resistances in the actual internal resistance sequence is recorded as the first offset error of the first electric curve; any internal resistance on the second electric curve is recorded as , x2 represents the acquisition time of the internal resistance, and the internal resistance at the time x2 is obtained on the first electrical curve, which is recorded as ,Will Recorded as the second prediction error; the average of the second prediction errors corresponding to all internal resistances on the second electric-varying curve is recorded as the second offset error of the first electric-varying curve; the ratio of the absolute value of the second prediction error to the absolute value of the first prediction error is recorded as the attention degree of the second electric-varying curve; The first offset error and the second offset error are fused according to the attention level of the second electric variable curve to obtain the offset error of the first electric variable curve; all internal resistances on the first electric variable curve are summed with the offset error to obtain a corrected first electric variable curve.

[0010] Preferably, the updating of the first electrical variation curve of each battery according to the first difference between F1 and F2 and the first duration from the current moment to time t1 includes the following specific steps: Obtaining a battery pack switching error based on a first difference between F1 and F2 and a first duration from the current moment to time t1; wherein the battery pack switching error is positively correlated with the first difference and negatively correlated with the first duration; When the battery pack switching error is greater than or equal to a first preset threshold, the first electrical change curve of each battery in the power supply group in the future time period T1 is re-obtained; when the battery pack switching error is less than the first preset threshold, the first electrical change curve is continued to be corrected based on the internal resistance of each battery in the power supply group that has been measured and the difference between the second electrical change curve and the first electrical change curve.

[0011] Preferably, the step of obtaining the distribution consistency of the standby groups in the target category obtained at any time and at adjacent time includes the following specific steps: For all target categories obtained at any moment and several other moments closest to the moment in time period T1; obtain the intersection and union ratio of the backup groups contained in any two target categories among all target categories, and the mean of the intersection and union ratio of all target categories is recorded as distribution consistency.

[0012] Preferably, the fusion of the first offset error and the second offset error according to the second electric curve attention level includes the following specific formula: ; Where Q1 represents the first offset error, Q2 represents the second offset error, w represents the normalized weight during fusion, w is positively correlated with the attention of the second electric-variable curve, and Q represents the offset error of the first electric-variable curve.

[0013] Preferably, the obtaining of the internal resistance growth rate at time i includes the following specific steps: Read the internal resistance of the first electrical curve at time i+y , y is the preset value; Denote the internal resistance growth rate at time i, and y0 is the preset scaling factor.

[0014] Other embodiments of the present invention provide a status monitoring device for substation batteries, which includes a plurality of batteries, a plurality of main control boards, and a server computer, wherein a main control board is installed on each battery; the main control board and the server computer include a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program runs the above-mentioned status monitoring method for substation batteries.

[0015] The beneficial effects of the technical solution of the present invention are: In the present invention, all batteries are divided into a power supply group and several backup groups. A first model is used to obtain a first electrical profile for each battery within a future time period T1. The first electrical profiles of all batteries within the power supply group are used to evaluate the health factor of the power supply group at each moment. This process provides an early estimate of the battery health by obtaining the first electrical profiles over a longer time scale in the future. Furthermore, a backup group switching time is obtained, at which the difference between the health factors of all backup groups and the health factor of the power supply group is maximized. The backup group switching time indicates the moment when the power supply group switches to the backup group.

[0016] Furthermore, at the current moment in time period T1, a second model is used to obtain a second electrical profile for each battery in a future time period T2. The first electrical profile is then corrected based on the measured internal resistance of each battery in the power supply group and the differences between the second electrical profile and the first electrical profile. This correction allows for real-time updates of the first electrical profile, preventing the long acquisition time required to update the battery status over longer timescales and, to a certain extent, minimizing offset errors in the first electrical profile.

[0017] Furthermore, the modified first power-varying curve is used to re-acquire the standby group switching time t1 and switching suitability F2; the first model has more parameters than the second model; the duration of T1 is greater than T2; the first power-varying curve of each battery is updated based on the first difference between F1 and F2 and the first duration from the current moment to time t1, and the power supply group is switched to the standby group at the standby group switching time t1. This process determines the subsequent update process of the first power-varying curve based on the change in switching suitability under the modified first power-varying curve and the time distance to switching the standby battery group, ensuring that after switching to the standby group at time t1, it can subsequently provide stable power supply for a long period of time, while the current power supply group can also continue to provide stable power supply for a long period of time; it also avoids the offset error in the standby group's first power-varying curve causing the magnitude of F2 to be submerged in the errors in F2 and F1, resulting in obvious errors in the switching suitability F2 and time t1, and ultimately causing the power grid to lose power for a short period of time after switching to the standby group at time t1.

[0018] The significance of long-term stable power supply lies in the following: Firstly, whether it is the standby group after the switchover or the current power group, its long-term stable power supply improves battery utilization and fully utilizes the battery performance, avoiding situations where the entire battery group cannot continue to supply power after a short period of use of one or more batteries. Secondly, long-term stability ensures that the first power curve can be repeatedly corrected in subsequent processes, further reducing the problem of offset errors in the first power curve (of the batteries in the power group) that may lead to large errors in the standby group switchover time and switching suitability. This ensures that the batteries can provide stable power for a long time before and after the switchover to the standby group. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] Figure 1This is a flowchart of the steps of a method for monitoring the status of a substation battery provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0021] To further illustrate the technical means and effectiveness of the present invention in achieving its intended objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation, structure, features, and effectiveness of the method and device for monitoring the condition of substation batteries proposed in accordance with the present invention. In the following description, references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0022] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0023] The specific scheme of the method and device for monitoring the status of substation batteries provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0024] Example 1: See also Figure 1 , which shows a flowchart of a method for monitoring the status of a substation battery according to an embodiment of the present invention, the method comprising the following steps: Step S101: All storage batteries are divided into a power supply group and several backup groups, and the status of each storage battery is monitored.

[0025] All batteries in the substation are divided into several groups (for example, 60 batteries form a group), and all batteries in each group are connected in series. One group is selected from all the groups to supply power to the substation bus (or the power grid). This group is recorded as the power supply group, and the other groups are used as backup. This is so that when the power supply of the power supply group is abnormal (for example, when the internal resistance changes abnormally), one group can be selected from the other groups to supply power. In this embodiment, the other backup groups are recorded as the backup group.

[0026] For any battery, a voltage sensor measures the voltage output by the battery at each moment, a current sensor measures the current drawn by the battery at each moment, and a temperature sensor measures the temperature of the battery at each moment. An AC injection sensor measures the internal resistance of the battery at each moment, and an integration method is used based on the voltage and current to obtain the remaining capacity at each moment. The specific method for obtaining the remaining capacity is well known and will not be described in detail in this embodiment.

[0027] In this embodiment, every 1 second is a moment.

[0028] Step S102: Use the first model to predict a first electrical variation curve of each battery in a future time period T1.

[0029] For each battery in the power supply group, for the first time period most recent before the current moment, at any moment in the first time period, the voltage, current, internal resistance, temperature, and power measured at that moment are regarded as a first vector; the first vectors at all moments in the first time period constitute a first vector sequence.

[0030] For each battery in the backup group, since the battery may not be in a working state at the current moment (for example, neither in a charging state nor in a discharging state), at this time, for the moment when the battery was last in a working state (charging state or discharging state), for the first time period most recently before that moment, at any moment in the first time period, the voltage, current, internal resistance, temperature, and power measured at that moment are regarded as a first vector; the first vectors at all moments in the first time period constitute a first vector sequence.

[0031] In this embodiment, it is taken into consideration that the change in the internal resistance of the battery is affected by many factors. For example, the current, voltage, temperature during the charge and discharge history, as well as overcharge and discharge conditions, can all cause changes in the internal resistance. Even when the battery is stored for a long time, the internal resistance may change.

[0032] Based on this, perform the following steps (1) and (2): (1) The first vector sequences of all batteries in the standby group and the first vector sequences of all batteries in the power group are sequentially input into the first model. The internal resistance of each battery at each moment in the future time period T1 is predicted. These internal resistances constitute the first electrical curve of the battery (the horizontal axis is the time, the vertical axis is the internal resistance). Note that the starting time of the horizontal axis of the first electrical curve obtained for each battery (i.e., the first horizontal axis) is the moment after the first model outputs the first electrical curve.

[0033] (2) After the first electrical variation curves of all batteries in all backup groups and power supply groups are obtained in sequence in step (1), step (1) is repeated every certain time interval (recorded as time L) to update the first electrical variation curves of all batteries.

[0034] First, it should be noted that in order to assess the health of the battery by analyzing the internal resistance over a longer time scale in the future, time period T1 in this embodiment represents a time period within one hour after the current moment, and the length of this time period is one hour. In other embodiments, time period T1 can be set to other values, which are not specifically limited in this embodiment. In addition, in this embodiment, time L is equal to 0.8 hours. In other embodiments, time L can be set to other values, preferably less than the length of time period T1.

[0035] It should be further noted that in this embodiment, the first time period is set to a relatively long duration to facilitate prediction of future internal resistance changes over a larger time scale based on historical changes in current, voltage, power, temperature, and internal resistance over a larger time scale. As an example, the first time period closest to the current moment refers to the time period within two hours prior to the current moment (including the current moment). In other embodiments, the first time period may be set to other time periods, and this embodiment does not specifically limit this.

[0036] In particular, if the battery has not been used for two hours before the current moment (for example, it has been in a long-term storage state), there are empty elements in the first vector sequence (the first vector cannot be collected at these empty elements). At this time, the current and voltage in the first vector at the empty elements are set to 0, and the power, internal resistance, and temperature are the power, internal resistance, and average temperature when the battery is in an idle state.

[0037] It should also be noted that in some embodiments, considering that the batteries in the standby group are not supplying power (perhaps in a charging state or a storage state), if the battery in the standby group has been in a storage state (i.e., neither charged nor discharged) for a short period of time (e.g., less than two days), its internal resistance is deemed unchanged. In this case, the internal resistance in the first electrical variation curve of the battery in the standby group is directly set to the internal resistance in the last obtained first electrical variation curve.

[0038] As an example, the first model is a Transformer network. Transformer networks are well-known technologies. For example, the GPT-3 / 4 model and the DEEPSEED model all utilize Transformer networks. It should be noted that although the GPT-3 / 4 model and the DEEPSEED model are used to process natural language, their principle is to convert natural language into a sequence of word vectors for processing. In this embodiment, the input is a sequence of first vectors, and the output is a sequence of internal resistance (i.e., the first electric curve). Therefore, in this embodiment, the GPT-3 / 4 model and the DEEPSEED model can be directly used as the first model. In other embodiments, the model structure of the GPT-3 / 4 model and the DEEPSEED model can be adjusted (for example, by deleting several self-attention layers) and then used as the first model, or the Transformer network can be reused to design the first model. This embodiment does not specifically limit this.

[0039] As an example, the training method of the first model is: For any battery in the charging or discharging state, obtain the current, voltage, power, temperature, and internal resistance collected at all times within 3 hours before any moment. The first vector sequence consisting of the current, voltage, power, temperature, and internal resistance collected in the first two hours is used as a sample, and the sequence consisting of the internal resistance collected in the last hour is recorded as the label.

[0040] A large number of samples and labels are obtained for the charge and discharge processes of all batteries, which constitute a data set. The first model is trained using a stochastic gradient descent algorithm and a mean square error loss function using the data set.

[0041] Since the specific training process of the Transformer network is well known, this embodiment will not be described in detail.

[0042] Step S103: Obtain the health factor of the power supply group at each moment according to the internal resistance value and internal resistance change rate at each moment in the first electrical variation curve of all batteries in the power supply group.

[0043] The health factor is used to describe the health status of all batteries in the power supply group; the larger the health factor, the better the health of all batteries in the power supply group and the longer they can continue to provide stable power to the grid; when the health factor is smaller, the health of all batteries in the power supply group is abnormal and there is a risk that they will not be able to continue to provide stable power to the grid for a long time.

[0044] As an example, the health factor of the power supply group at each moment is obtained based on the internal resistance value and internal resistance change rate of each moment in the first electrical variation curve of all batteries in the power supply group, including: For the first electrical curve of any battery in the power supply group, read the internal resistance at time i in the first electrical curve , read the internal resistance at time i+y , this embodiment is described by taking y=600 as an example. Recorded as the internal resistance growth rate at time i, the faster the internal resistance from time i to time i+y grows, The bigger. Indicates the preset scaling factor. In this embodiment, For example, y0 can be set to other values ​​in other embodiments. This embodiment does not specifically limit it. Its purpose is to remove Dimension of .

[0045] Specifically, if the first electrical variation curve does not have time i+y, this embodiment sets the internal resistance at time i+y on the first electrical variation curve to the internal resistance at the last time on the first electrical variation curve. In other embodiments, a linear interpolation algorithm may also be used to interpolate the internal resistance of the first electrical variation curve at time i+y.

[0046] This Implementation Order .in represents the health index at time i, exp() represents the exponential function with a natural constant as the base, It is a preset resistance value used to remove the dimension. In this embodiment, R0 is equal to the average of the initial internal resistance of all batteries in the power supply group (that is, the internal resistance when leaving the factory). In other embodiments, R0 can also be set to other resistance values, which is not specifically limited in this embodiment.

[0047] In other examples, you can .

[0048] The greater the internal resistance growth rate or the greater the internal resistance at time i, the smaller the health index, indicating that the battery is less likely to continue to supply power to the grid for a long time at time i and thereafter. The smaller the internal resistance growth rate or the smaller the internal resistance at time i, the larger the health index, indicating that the battery is more likely to continue to supply power to the grid for a long time at time i and thereafter.

[0049] At this point, each battery in the power supply group obtains a corresponding health indicator at time i.

[0050] Specifically, when the health indicator at the i-th moment is less than or equal to the health threshold, it indicates that the battery is no longer able to operate, and the health indicator is set to 0. This embodiment uses the health threshold equal to 0.3 as an example. Other embodiments may set it to other values, which are not specifically limited in this embodiment.

[0051] This embodiment takes into account that if even one battery in a power supply group experiences a health anomaly, the entire group may be unable to provide power to the grid for an extended period (also known as the "barrel effect"). Therefore, whenever a first power curve for any battery in the power supply group is obtained, the minimum value of the health indicators corresponding to all first power curves obtained for the power supply group at time j is used as the health factor of the power supply group at that time.

[0052] Specifically, if a first electrical profile does not have a corresponding health indicator at time j, in this embodiment, the health indicator of the first electrical profile at time j is set to the health indicator at the last time in the first electrical profile. In other embodiments, a linear interpolation algorithm can also be used to interpolate the health indicator of the first electrical profile at time j.

[0053] In summary, whenever the first electrical change curve of any battery in the power supply group is obtained, for any moment (for example, moment j) on the first electrical change curve, this embodiment obtains the health factor of the power supply group at each moment (for example, moment j) based on the first electrical change curves of all batteries in the power supply group.

[0054] Step S104: Obtain the standby group switching time, at which the difference between the health coefficients of all standby groups and the health coefficient of the power supply group is the largest, and the maximum value is recorded as the switching suitability F1.

[0055] The process for obtaining the health coefficient of each standby group at each moment is similar to that of step S103. The general process is as follows: after obtaining the first power curve of any battery in each standby group, at any moment on that first power curve, the health coefficient of the standby group at that moment is obtained based on the first power curves of all batteries in the standby group. It should be noted that in this embodiment, the health coefficient of each standby group at each moment is obtained first, and then the health coefficient of the power supply group at each moment is obtained.

[0056] Obtain the standby group switching time, denoted as t. At time t, the difference between the health coefficients of all standby groups and the health coefficient of the power supply group is the largest. The maximum value is recorded as the switching suitability, expressed as F1. Among them, at time t, the standby group that makes the difference in health coefficient the largest is recorded as the target group.

[0057] The standby group switching time (i.e., time t) indicates that the batteries in the target group need to be used for power supply at time t. The batteries in the original power supply group may not be able to provide stable power to the grid for a long time and need to be repaired and replaced as the standby group.

[0058] The greater the switching suitability, the less likely the power supply group will be able to provide long-term stable power supply to the grid at time t. However, if there is a standby group (i.e., the target group) at time t, when the target group takes over power supply from the power supply group, the grid can be provided with long-term stable power supply more reliably.

[0059] Step S105: In the time period T1, use the second model to predict the second electrical variation curve of each battery in the future time period T2.

[0060] When step S102 predicts the first electrical profile of each battery at a relatively large time scale, it takes a long time (in this embodiment, a single run of the first model takes 10 seconds). This is because the first model has a large number of parameters and a large amount of input and output data, which makes it impossible to obtain the first electrical profiles of all batteries in the standby group and the power supply group at the same time. It is also impossible to update the predicted first electrical profiles in real time when predicting at a relatively large time scale (i.e., it is impossible to update the first electrical profiles of all batteries in the power supply group at every moment). As described in (2) of step S102, the first electrical profiles of all batteries can only be updated once every certain time interval (i.e., time L).

[0061] Based on this, it can be seen that after any battery in the power supply group obtains a first electrical profile using the first model and before obtaining its first electrical profile again using the first model, the first electrical profile cannot be updated or corrected. This can lead to an offset error in the first electrical profile (i.e., the trend of the first electrical profile gradually deviates from the actual internal resistance of the battery). This offset error interferes with the accuracy of the standby group switching timing in step S104. This can result in a delay in switching to the standby group, which can lead to further deterioration of the power supply group's health (consistently increasing internal resistance) or, in worse cases, failure to maintain grid stability.

[0062] In this embodiment, for any battery in the power supply group, the battery corresponds to a first electrical curve in a time period T1. For each moment in the time period T1 corresponding to the battery, the following processing is performed: Each moment in the time period T1 is recorded as a target moment. At the target moment, the second model is used to predict the second power curve of the battery in the future time period T2, specifically including: For the second time period closest to the target time, at any time in the second time period, the voltage, current, internal resistance, temperature, and power measured at that time are regarded as a second vector; the second vectors at all times in the second time period constitute a second vector sequence.

[0063] The second vector sequence is input into the second model, and the internal resistance at all moments in the future time period T2 is output. These internal resistances constitute the second electrical variation curve (the horizontal axis is the moment, the vertical axis is the internal resistance, and the starting moment of the horizontal axis is recorded as the moment after the target moment).

[0064] In this embodiment, the most recent second time period refers to the time period within 10 minutes before the target time (including the target time). In other embodiments, the second time period can be set to other time periods. This embodiment does not make specific limitations, and it is only necessary to ensure that the length of the second time period is less than the length of the first time period.

[0065] The future time period T2 in this embodiment represents a time period within 1 minute after the target time. In other embodiments, the time period T2 can be set to other values. This embodiment does not make any specific restrictions. It only needs to ensure that the length of the time period T2 is less than the length of the time period T1.

[0066] The second model has fewer parameters than the first model and predicts a second electrical profile over a short timescale. In this embodiment, the second model's parameters are one percent of those of the first model. Due to the small number of parameters in this embodiment, the small number of outputs, and the high speed of operation, the second electrical profile of each battery in the power supply group can be obtained at each moment within time period T2. This second electrical profile is free of long-term offset errors (although errors may exist, manifesting as fluctuations in the internal resistance over a local timeframe, referred to as fluctuation errors).

[0067] As an optional example, the second model is an LSTM neural network.

[0068] As a preferred example, the second model is a Transformer network.

[0069] As an example, the training method of the second model is: For any battery in the charging or discharging state, obtain the current, voltage, charge, temperature, and internal resistance collected at all times within 11 minutes before any time. The first vector sequence consisting of the current, voltage, charge, temperature, and internal resistance collected within 10 minutes is used as a sample, and the sequence consisting of the internal resistance collected in the last minute is recorded as the label.

[0070] A large number of samples and labels are obtained for the charge and discharge processes of all batteries, which constitute a dataset. The second model is trained using a stochastic gradient descent algorithm and a mean square error loss function using the dataset.

[0071] In some other examples, a time series forecasting model such as ARIMA or Prophet is used as the second model. In this case, the second model only needs to input a sequence consisting of all internal resistances in the second vector sequence.

[0072] Step S106: Correct the first electrical variation curve based on the difference between the measured internal resistance of all batteries in the power supply group and the first electrical variation curve, and the difference between the second electrical variation curve and the first electrical variation curve.

[0073] For any battery in the power supply group, the battery corresponds to a first electrical profile within a time period T1. During this time period T1, the battery uses a sensor to collect internal resistance at each moment, while also using a second model to predict a second electrical profile for the battery (see step S105 for details). Because the first electrical profile is predicted over a larger time scale, its error manifests as an offset over that larger time scale. Because the second electrical profile is predicted over a smaller time scale, its error manifests as a localized, short-term fluctuation.

[0074] During the time period T1, as time passes, in this embodiment, whenever the battery obtains a second electrical curve at each moment, the first electrical curve is corrected by using the difference between the internal resistance measured during the time period T1 (i.e., the internal resistance collected by the sensor using the AC injection method) and the first electrical curve, as well as the difference between the second electrical curve and the first electrical curve.

[0075] Although this embodiment cannot obtain or update the first electrical variation curve in real time (i.e., at every other moment) by using the first model, it can use the second model to update or correct the first electrical variation curve in real time, thereby avoiding the problem of being unable to obtain an accurate standby group switching time due to an offset error in the first electrical variation curve.

[0076] For all the batteries in the power supply group, in the time period T1 corresponding to each of these batteries, the first power curve is corrected in real time using the second power curve of each battery.

[0077] Step S107: Use the corrected first power change curve to re-obtain the standby group switching time t1 and the switching suitability F2; update the first power change curve of each battery according to the first difference between F1 and F2, and the first duration from the current moment to time t1, and at the same time, switch the power supply group to the standby group at the standby group switching time t1.

[0078] To summarize the above steps: for each battery in the power supply group, after obtaining the first electrical curve in time period T1, the first electrical curve is updated by real-time prediction of the second electrical curve in time period T1 to reduce the offset error of the first electrical curve.

[0079] However, it should be noted that when correcting the first electric variable curve, only the overall error offset of the first electric variable curve can be corrected, which cannot further correct the local change trend on the first electric variable curve, resulting in a certain degree of offset error in the corrected first electric variable curve.

[0080] However, the purpose of battery status monitoring in this embodiment is not to obtain an accurate first power-varying curve. Instead, it is necessary to use the first power-varying curve to accurately and reliably determine the standby group switchover time, so that the power supply group can be switched to the standby group (also known as the target group) at the standby group switchover time. Therefore, when updating and revising the first power-varying curve in this embodiment, it is sufficient to ensure the accuracy and reliability of the standby group switchover time. This accuracy and reliability specifically means that, after switching to the standby group at the standby group switchover time, the power supply group can avoid the problem of being unable to provide stable power for a long time, and the standby group can also take over the power supply group's stable power supply for a long time.

[0081] Based on this, the first power curves of all batteries in the power supply group are corrected at each moment within their corresponding time period T1. This embodiment uses the corrected first power curves to re-acquire the standby group switching time and switching suitability, as well as the target group. The specific process is similar to step S104 and will not be further described here. The standby group switching time and switching suitability obtained at this point are denoted as t1 and F2, respectively.

[0082] That is, each time the first electrical variation curves of all batteries are corrected, a standby group switching time t1 and a switching suitability F2 can be obtained.

[0083] At the current moment in time period T1 (note that as time goes by, the current moment is different from the current moment described in step S102), the difference between F1 and F2 is recorded as the first difference, and the interval time between the current moment and time t1 is recorded as the first duration.

[0084] In some preferred examples, the ratio of the difference between F1 and F2 to F1 is recorded as the first difference. Specifically, in some embodiments, to prevent the denominator from being zero, a smaller value (e.g., 0.01) is automatically added to the denominator. Other embodiments may use other methods to prevent the denominator from being equal to zero. For ease of description, this embodiment does not consider the case where the denominator is zero.

[0085] In some preferred examples, the ratio of the interval time between the current moment and time t1 to the length of time period T1 is recorded as the first duration.

[0086] Furthermore, this embodiment obtains a battery switching error based on the first difference and the first duration. The battery switching error is positively correlated with the first difference and negatively correlated with the first duration. That is, the larger the first difference, the larger the battery switching error, and the smaller the first difference, the smaller the battery switching error. The longer the first duration, the smaller the battery switching error, and the smaller the first duration, the larger the battery switching error.

[0087] As a preferred example, the battery pack switching error C=C1 / C2, where C1 represents the first difference and C2 represents the first duration.

[0088] As an optional example, the battery pack switching error C=C1-C2.

[0089] The battery pack switching error describes the suitability of switching to the backup group (also known as the target group) at time t1. The smaller the battery pack switching error, the more suitable the switch to the backup group (also known as the target group) at time t1. Specifically, when the first difference is smaller or the first duration is longer (the battery pack switching error is smaller), it indicates that the switching suitability F2 of the modified first electrical variation curve has not significantly decreased (or may even increase). In this case, switching to the backup group (also known as the target group) at time t1 is more likely to ensure subsequent long-term stable power supply. Furthermore, since the current moment is still far from time t1, the current power supply group can also continue to provide stable power for a long time.

[0090] Conversely, the larger the battery pack switching error, the less suitable it is to switch to the backup group (also known as the target group) at time t1. Specifically, when the first difference is larger or the first duration is smaller (in this case, the battery pack switching error is larger), the switching suitability F2 of the corrected first electrical profile is significantly reduced. In this case, it is difficult to ensure subsequent long-term stable power supply after switching to the backup group (target group) at time t1. Furthermore, considering that the offset error in the backup group's first electrical profile can cause errors in the switching suitability F2 and F1, when F2 is significantly reduced, the error between F2 and F1 exceeds the magnitude of F2 (i.e., the magnitude of F2 is submerged in the error between F2 and F1), resulting in significant errors in the switching suitability F2 and time t1. This can lead to a short-term power outage after switching to the backup group at time t1. Furthermore, since the current moment is close to time t1, the current power supply group cannot continue to provide power to the grid for an extended period of time.

[0091] It should be noted that the significance of long-term stable power supply lies in the following: on the one hand, whether it is the standby group after switching or the current power group, its long-term stable power supply improves battery utilization and fully utilizes the battery performance, avoiding the situation where the entire battery group cannot continue to supply power after a short period of use of one or more batteries. On the other hand, long-term stability ensures that the first power curve can be corrected multiple times in the subsequent process, further reducing the problem of offset errors in the first power curve (of the batteries in the power group) that may lead to large errors in the standby group switching time and switching suitability. As a result, the batteries can provide long-term stable power supply before and after the subsequent switch to the standby group.

[0092] Based on this, in this embodiment, when the battery group switching error is less than the first prediction threshold th1, continuing to use the above method to update the first power change curve can ensure that the battery can provide stable power for a long time before and after switching to the backup group. Subsequently, the above method is continued to be used to correct the first power change curve (i.e., the first power change curve is corrected using the second power change curve). If the backup group switching time t1 is reached at the current moment, the power supply group is switched to the backup group (i.e., the target group).

[0093] Conversely, if the battery pack switching error is greater than or equal to the first obtained threshold th1, continuing to use the above method to update the first power curve may not guarantee long-term stable power supply from the battery before and after switching to the backup pack. In this case, the first model is immediately re-used to sequentially predict the first power curve for each battery in the power pack for the future time period T1. This updates the first power curve and avoids the potential problem of long-term stable power supply failure when continuing to use the second power curve to correct the first power curve.

[0094] This embodiment is described by taking th1=0.2 as an example. In other embodiments, it can be set to other values, which is not specifically limited in this embodiment.

[0095] After reusing the first model to predict the first power profile for each battery in the power supply group during the future time period T1, the method described above should be used to continue obtaining the backup group switch time and updating the first power profile. Note that if the backup group switch time has already been reached, the power supply group will be switched to the backup group (also known as the target group).

[0096] This concludes the present embodiment.

[0097] In summary, in this embodiment, all batteries are divided into a power supply group and several backup groups. A first model is used to predict the first electrical profile of each battery within a future time period T1. The first electrical profiles of all batteries within the power supply group are used to assess the health factor of the power supply group at each moment. This process predicts the first electrical profiles over a longer time scale in the future to provide an early estimate of the battery health. Furthermore, a backup group switchover time is obtained, at which the difference between the health factors of all backup groups and the health factor of the power supply group is maximized. The backup group switchover time indicates the moment when the power supply group switches to the backup group.

[0098] Furthermore, at the current moment in time period T1, the second model is used to predict the second electrical profile of each battery in the future time period T2. The first electrical profile is then corrected based on the measured internal resistance of each battery in the power supply group and the difference between the second electrical profile and the first electrical profile. This correction allows for real-time updates of the first electrical profile, avoiding situations where battery status predictions over longer timescales take too long to be updated in real time, and to some extent, minimizing offset errors in the first electrical profile.

[0099] Furthermore, the modified first power-varying curve is used to re-acquire the standby group switching time t1 and switching suitability F2; the first model has more parameters than the second model; the duration of T1 is greater than T2; the first power-varying curve of each battery is updated based on the first difference between F1 and F2 and the first duration from the current moment to time t1, and the power supply group is switched to the standby group at the standby group switching time t1. This process determines the subsequent update process of the first power-varying curve based on the change in switching suitability under the modified first power-varying curve and the time distance to switching the standby battery group, ensuring that after switching to the standby group at time t1, it can subsequently provide stable power supply for a long period of time, while the current power supply group can also continue to provide stable power supply for a long period of time; it also avoids the offset error in the standby group's first power-varying curve causing the magnitude of F2 to be submerged in the errors in F2 and F1, resulting in obvious errors in the switching suitability F2 and time t1, and ultimately causing the power grid to lose power for a short period of time after switching to the standby group at time t1.

[0100] Example 2: As an optional example, in step S104 of the first embodiment, the method for obtaining the standby group switching time is: After obtaining the first electrical profile of each battery in the power supply group, for each moment in time period T1 (or in other words, for each moment in the first electrical profile), the difference between the health coefficient of any standby group and the health coefficient of the power supply group at that moment is obtained. This difference is recorded as the health difference of any standby group at that moment. The standby group with the largest health difference at that moment is recorded as the candidate group. In this case, each moment corresponds to a candidate group and a health difference. For all candidate groups at each moment in time period T1, the moment with the largest health difference among the candidate groups is obtained. This is recorded as the standby group switching time, denoted as t. The candidate group with the largest health difference is recorded as the target group. The maximum health difference is recorded as the switching suitability, denoted by F1.

[0101] As a preferred example, in step S104 of the first embodiment, the method for obtaining the standby group switching time is: The following processing is performed for each moment in the time period T1 (or for each moment in the first electrical variation curve): Follow the above optional example to obtain the health difference of any alternative group at each time. Perform K-means clustering on the health differences of all alternative groups at each time, clustering them into two categories. The mean of all health differences within each category is recorded as the cluster center, and the category with the largest cluster center is recorded as the target category.

[0102] For a target category obtained at any moment, obtain the moment and several (eg, 10) other moments closest to the moment in time period T1, and obtain the distribution consistency of the backup groups in the target categories at these moments.

[0103] Distribution consistency is determined by calculating the intersection-over-union ratio (IoU) of the backup groups within any two target categories (the ratio of the number of identical backup groups within the two target categories to the total number of backup groups). The average of these IoUs for all target categories within the target categories at these times is considered distribution consistency. Greater distribution consistency indicates a consistent distribution of backup groups within the target categories obtained at different times within a local timeframe.

[0104] The product of the cluster center of the target category at any moment and its distribution consistency is recorded as the second index at each moment. The moment when the second index is maximized is recorded as the standby group switching time t. The standby group with the largest health difference at standby group switching time t is the target group. The maximum health difference is recorded as the switching suitability, denoted by F1.

[0105] The standby group switching time t in this preferred example means that there is a standby group with a large health difference. At the same time, the standby group with a large health difference does not change significantly within a local time, indicating that after switching to the target group at the standby group switching time t, it can continue to provide stable power supply for a long time, and preliminarily avoid the situation where the standby group switching time t and the target group are inaccurate due to errors in the first electrical change curve of the battery.

[0106] As an optional example, in step S106 of the first embodiment, the first electrical variation curve is corrected based on the difference between the measured internal resistance of all batteries in the power supply group and the first electrical variation curve, and the difference between the second electrical variation curve and the first electrical variation curve, including the following method: For the target moment in the time period T1, the internal resistance collected by the sensor at all moments before the target moment (including the target moment) in the time period T1 is obtained. These internal resistances constitute the actual internal resistance sequence. For any internal resistance in the actual internal resistance sequence and the second electric variable curve, the collection time of the internal resistance is recorded as x, and the internal resistance is recorded as , obtain the internal resistance at time x on the first electrical curve, recorded as ,Will It is recorded as the first prediction error.

[0107] At this point, any internal resistance in the actual internal resistance sequence and the second electrical variation curve corresponds to a first prediction error. The average of the first prediction errors corresponding to all internal resistances in the actual internal resistance sequence and the second electrical variation curve is recorded as the offset error of the first electrical variation curve. The offset error is summed with all internal resistances in the first electrical variation curve to obtain the corrected first electrical variation curve.

[0108] As a preferred example, in step S106 of the first embodiment, the first electrical variation curve is corrected based on the difference between the measured internal resistance of all batteries in the power supply group and the first electrical variation curve, and the difference between the second electrical variation curve and the first electrical variation curve, including the following method: This example takes into account that the second electrical variation curve also has errors, which results in the possibility that the corrected first electrical variation curve obtained at the above optional moment may still have obvious offset errors.

[0109] In this example, any internal resistance in the actual internal resistance sequence is recorded as , x1 represents the acquisition time of the internal resistance, and the internal resistance at the time x1 is obtained on the first electrical curve, which is recorded as ,Will It is recorded as the first prediction error.

[0110] At this point, any internal resistance in the actual internal resistance sequence corresponds to a first prediction error, and the average of the first prediction errors corresponding to all internal resistances in the actual internal resistance sequence is recorded as the first offset error of the first electrical variation curve.

[0111] Similarly, any internal resistance on the second electric curve is recorded as , x2 represents the acquisition time of the internal resistance, and the internal resistance at the time x2 is obtained on the first electrical curve, which is recorded as ,Will It is recorded as the second prediction error.

[0112] At this point, any internal resistance on the second electrical variation curve corresponds to a second prediction error, and the average of the second prediction errors corresponding to all internal resistances on the second electrical variation curve is recorded as the second offset error of the first electrical variation curve.

[0113] In this embodiment, the ratio of the absolute value of the second prediction error to the absolute value of the first prediction error is recorded as the second electrical variation curve attention level.

[0114] In some other embodiments, the inverse of the ratio of the absolute values ​​of the first prediction errors may also be recorded as the second electrical variation curve attention level.

[0115] The greater the attention paid to the second electric-variable curve, the more the first electric-variable curve conforms to the actually collected internal resistance variation trend. Therefore, even if more attention is paid to the future internal resistance variation trend represented by the second offset error, the first internal resistance will not be significantly offset. The significance of paying more attention to the future internal resistance variation trend represented by the second offset error is that a longer time length can be used to correct the first electric-variable curve.

[0116] The lower the attention level of the second electric-variable curve, the more the first electric-variable curve deviates from the actual acquired internal resistance variation trend. In this case, it is more necessary to use the actual acquired internal resistance to correct the first electric-variable curve, and the less necessary to consider the future internal resistance variation trend represented by the second offset error, so as to avoid making the offset error of the first electric-variable curve more obvious. In addition, (for example, when the attention level of the second electric-variable curve approaches 0), only the actual internal resistance sequence is considered, and the future internal resistance variation trend represented by the second offset error is no longer considered.

[0117] Furthermore, the first offset error and the second offset error are fused according to the attention level of the second electric variation curve to obtain the offset error of the first electric variation curve. All internal resistances on the first electric variation curve are summed with the offset error to obtain a corrected first electric variation curve.

[0118] As an example, the first offset error and the second offset error are fused according to the attention level of the second electric variable curve to obtain the offset error of the first electric variable curve, including the formula: ; in represents the first offset error, represents the second offset error, represents the weight during fusion, w=1-exp(-q), q represents the attention of the second electric-variable curve, where the formula 1-exp(-q) is used to normalize q, exp() represents an exponential function with a natural constant as the base, and Q represents the offset error of the first electric-variable curve.

[0119] Specifically, in this embodiment, when q is less than 0.1, q is set to 0; when q is greater than 3, q ​​is set to 3. In other embodiments, the value range of q may be limited by other methods, which are not specifically limited in this embodiment.

[0120] Example 3: This embodiment provides a condition monitoring system for substation batteries, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor; when the processor executes the computer program, the above-mentioned condition monitoring method for substation batteries is implemented.

[0121] Example 4: This embodiment provides a condition monitoring device for substation batteries. The device includes several batteries, each of which is equipped with a set of sensors, including: a voltage sensor, a current sensor, a temperature sensor (installed on the battery surface), and an AC injection method sensor (such as the Fluke BT500 series), which are used to measure data such as the battery's voltage, current, temperature, and internal resistance.

[0122] In addition, each battery is also equipped with a corresponding main control board (such as an ESP32 series microcontroller). The above sensors are all installed on this main control board.

[0123] The device further includes a server computer which is connected to and communicates with the main control boards of all batteries via a WIFI module (or via a network cable in other embodiments).

[0124] All main control boards and server computers contain memory and processors for running the condition monitoring system for substation batteries.

[0125] Specifically, when executing the computer program, all main control boards are used to collect data such as voltage, current, temperature, internal resistance, and power. Furthermore, the second model described in Example 1 runs on each main control board, and step S105 of Example 1 is executed on the main control board. Because the first model requires a large number of parameters, it runs on the server computer. When the computer program is executed on the server computer, the remaining methods of Example 1 are implemented.

[0126] Embodiment 5: This embodiment makes the following adjustments to the method in embodiment 1: The first adjustment method is to increase the interval between each moment, for example, every 3 seconds is a moment.

[0127] This process takes into account that the capabilities of running computer programs vary when using different main control boards and server computers. In particular, when running the first model, the server computer uses different GPUs and has different running speeds. This embodiment reduces the amount of data by increasing the interval between each moment and reducing the number of sensor sampling times to adapt to the computing capabilities of the main control board and the server computer.

[0128] The second adjustment method is: the time period within 5 hours after the current time is used as time period T1. The time period within 10 hours before the previous time is used as the first time period. y in step S103 is increased by 2 times.

[0129] This process increases the time scale of time period T1, the first time period, and the parameter y to ensure that the internal resistance changes over a longer period of time can be described, avoiding the situation where the battery internal resistance changes are not obvious over a shorter period of time, and ensuring that abnormal changes in internal resistance (that is, the health factor gradually decreases) can be discovered in a timely and early manner.

[0130] The third adjustment method is to immediately switch to the backup group (e.g., the target group described in Example 1, or randomly switch to the backup group) when faced with a special situation. Special situations include: batteries in the power supply group experiencing a short circuit (e.g., a current exceeding the rated battery current), low battery charge (e.g., less than 10% of total capacity), or high temperature (e.g., greater than 120% of the rated operating temperature).

[0131] The fourth adjustment method is: use a controllable selective switch to switch the power supply group to the standby group. It should be noted that after switching to the standby group, the original power supply group needs to be charged, repaired and maintained before it can continue to be used as a standby group.

[0132] The fifth adjustment method is: displaying the voltage, current, internal resistance and other data collected in real time from each battery on the screen, and also displaying the corrected first electrical curve on the screen in real time to achieve visual monitoring.

[0133] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for monitoring the status of a battery in a substation, characterized in that: The method comprises the following steps: All batteries are divided into a power supply group and several backup groups; a first electrical change curve of each battery in a future time period T1 is obtained; the first electrical change curves of all batteries in the power supply group are used to evaluate the health factor of the power supply group at each moment and obtain the backup group switching time, where the difference between the health factors of all backup groups and the health factor of the power supply group is the largest at the backup group switching time, and the maximum value is recorded as the switching suitability F1; at the current moment in time period T1, a second electrical change curve of each battery in a future time period T2 is obtained; the first electrical change curve is corrected based on the internal resistance and the difference between the second electrical change curve and the first electrical change curve of each battery in the power supply group, and the backup group switching time t1 and the switching suitability F2 are re-obtained using the corrected first electrical change curve; the duration of T1 is greater than T2; the first electrical change curve of each battery is updated based on the first difference between F1 and F2 and the first duration from the current moment to time t1, and the power supply group is switched to the backup group at the backup group switching time t1.

2. The method for monitoring the status of a substation battery according to claim 1, characterized in that: The specific steps of the health coefficient of the power supply group at each moment are as follows: For the first electrical curve of any battery in the power supply group, read the internal resistance at time i in the first electrical curve , obtain the internal resistance growth rate at time i; the health index of the first electric curve at time i and the internal resistance growth rate at time i, were negatively correlated; Whenever the first electrical variation curve of any battery in the power supply group is obtained, for each moment on the first electrical variation curve, and for all the first electrical variation curves obtained in the power supply group, the minimum value of the health indicators of all the first electrical variation curves at each moment is used as the health coefficient of the power supply group at each moment.

3. The method for monitoring the status of a substation battery according to claim 1, characterized in that: The step of obtaining the standby group switching time, wherein the difference between the health coefficients of all standby groups and the health coefficient of the power supply group is the largest at the standby group switching time, and the maximum value is recorded as the switching suitability F1, includes the following specific steps: Whenever the first power curve of each battery in the power supply group is obtained, for the time period T1 in which the first power curve is located, and for any moment in time period T1, the difference between the health coefficient of any standby group and the health coefficient of the power supply group at that moment is obtained, and recorded as the health difference of any standby group at that moment; Perform K-means clustering on the health differences of all backup groups at each moment and cluster them into two categories. The mean of all health differences in each category is recorded as the cluster center, and the category with the largest cluster center is recorded as the target category. Obtain the distribution consistency of the backup groups in the target category obtained at any moment and adjacent moments; record the product of the cluster center of the target category at any moment and the distribution consistency as the second index at each moment; within time period T1, record the moment with the maximum second index as the backup group switching time, recorded as t, and select the backup group with the largest health difference at the backup group switching time t as the target group; The maximum health difference is recorded as the switching fitness, represented by F1.

4. The method for monitoring the status of a substation battery according to claim 1, characterized in that: The first electric change curve is corrected according to the difference between the measured internal resistance of each battery in the power supply group and the second electric change curve and the first electric change curve, and the specific steps include the following: Any moment in the time period T1 is recorded as the target moment, and the internal resistance collected at all moments before the target moment in the time period T1 constitutes the actual internal resistance sequence; Any internal resistance in the actual internal resistance sequence is recorded as , x1 represents the acquisition time of the internal resistance, and the internal resistance at the time x1 is obtained on the first electrical curve, which is recorded as ,Will Recorded as the first prediction error; the average of the first prediction errors corresponding to all internal resistances in the actual internal resistance sequence is recorded as the first offset error of the first electric curve; any internal resistance on the second electric curve is recorded as , x2 represents the acquisition time of the internal resistance, and the internal resistance at the time x2 is obtained on the first electrical curve, which is recorded as ,Will Recorded as the second prediction error; the average of the second prediction errors corresponding to all internal resistances on the second electric-varying curve is recorded as the second offset error of the first electric-varying curve; the ratio of the absolute value of the second prediction error to the absolute value of the first prediction error is recorded as the attention degree of the second electric-varying curve; The first offset error and the second offset error are fused according to the attention level of the second electric variable curve to obtain the offset error of the first electric variable curve; all internal resistances on the first electric variable curve are summed with the offset error to obtain a corrected first electric variable curve.

5. The method for monitoring the status of a substation battery according to claim 1, characterized in that: The updating of the first electrical change curve of each battery according to the first difference between F1 and F2 and the first time interval from the current moment to time t1 includes the following specific steps: Obtaining a battery pack switching error based on a first difference between F1 and F2 and a first duration from the current moment to time t1; wherein the battery pack switching error is positively correlated with the first difference and negatively correlated with the first duration; When the battery pack switching error is greater than or equal to a first preset threshold, the first electrical change curve of each battery in the power supply group in the future time period T1 is re-obtained; when the battery pack switching error is less than the first preset threshold, the first electrical change curve is continued to be corrected based on the internal resistance of each battery in the power supply group that has been measured and the difference between the second electrical change curve and the first electrical change curve.

6. The method for monitoring the status of a substation battery according to claim 3, characterized in that: The specific steps of obtaining the distribution consistency of the standby groups in the target category obtained at any time and at adjacent time are as follows: For all target categories obtained at any moment and several other moments closest to the moment in time period T1; obtain the intersection and union ratio of the backup groups contained in any two target categories among all target categories, and the mean of the intersection and union ratio of all target categories is recorded as distribution consistency.

7. The method for monitoring the status of a substation battery according to claim 4, characterized in that: The specific formula for fusing the first offset error and the second offset error according to the second electric curve attention is as follows: ； Where Q1 represents the first offset error, Q2 represents the second offset error, w represents the normalized weight during fusion, w is positively correlated with the attention of the second electric-variable curve, and Q represents the offset error of the first electric-variable curve.

8. The method for monitoring the status of a substation battery according to claim 2, characterized in that: The specific steps of obtaining the internal resistance growth rate at time i are as follows: Read the internal resistance of the first electrical curve at time i+y , y is the preset value; Denote the internal resistance growth rate at time i, and y0 is the preset scaling factor.

9. A condition monitoring device for storage batteries in a substation, comprising a plurality of storage batteries, a plurality of main control boards, and a server computer, wherein each storage battery is equipped with a main control board; the main control board and the server computer comprise a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The computer program runs the condition monitoring method for substation batteries according to any one of claims 1 to 8.

Images