Hybrid energy storage device of super capacitor and lithium battery for thermal power unit frequency modulation

Abstract

The application relates to the technical field of electric power energy storage, and discloses a super capacitor and lithium battery hybrid energy storage device for frequency modulation of a thermal power unit. The device comprises a decomposition module, an allocation module, a correction module and a processing module. The decomposition module decomposes an AGC frequency modulation instruction to obtain an initial boundary frequency; the allocation module allocates the initial boundary frequency to obtain an output instruction; the correction module combines current and temperature data to predict a temperature trend and correct the boundary frequency; and the processing module reallocates the boundary frequency to obtain a target output instruction. The application improves the matching degree between the power allocation result of the super capacitor and the lithium battery in the frequency modulation process of the thermal power unit and the real available power boundary of the two types of energy storage units.

Description

Technical Field

[0001] This application relates to the field of power energy storage technology, and in particular to a hybrid energy storage device of supercapacitor and lithium battery for frequency regulation of thermal power units. Background Technology

[0002] Currently, thermal power units participating in grid AGC frequency regulation generally employ a hybrid energy storage system composed of supercapacitors and lithium batteries to assist in completing the frequency regulation task. Existing technologies typically first perform spectrum decomposition processing on the AGC frequency regulation commands issued by the grid, dividing the frequency regulation commands into high-frequency and low-frequency components according to a boundary frequency. The high-frequency components are handled by supercapacitors with fast response speeds, while the low-frequency components are handled by lithium batteries with high energy density. Some existing technologies also combine the state of charge of the supercapacitors to predict the risk of power exceeding limits in the future, and adjust the boundary frequency or power allocation parameters in advance accordingly to reduce the frequency of lithium battery operation.

[0003] However, existing technologies, when determining boundary frequencies or adjusting allocation parameters, generally treat the maximum allowable charge and discharge power of supercapacitors and lithium batteries as fixed values ​​that do not change with operating conditions, only predicting and judging the trend of changes in the state of charge in terms of energy quantity. The equivalent capacitance and internal resistance of supercapacitors, as well as the maximum allowable charge and discharge current of lithium batteries, actually change significantly with temperature. The temperature inside the energy storage chamber also fluctuates considerably depending on the frequency of continuous frequency regulation, season, and environmental conditions. Existing technologies do not consider this factor, resulting in a deviation between the power boundary on which the risk of exceeding limits is based and the actual usable power boundary of the energy storage unit at the current temperature. This deviation gradually accumulates with the continuous frequency regulation. Summary of the Invention

[0004] This application provides a hybrid energy storage device of supercapacitor and lithium battery for frequency regulation of thermal power units, which solves the problem that the power allocation decision in the prior art does not consider the dynamic impact of the temperature change of the energy storage unit on its power boundary, and improves the matching degree between the power allocation result of supercapacitor and lithium battery and the actual usable power boundary of the two types of energy storage units during the frequency regulation of thermal power units.

[0005] This application provides a hybrid energy storage device combining a supercapacitor and a lithium battery for frequency regulation of a thermal power unit. The hybrid energy storage device includes: The decomposition module is used to perform spectrum decomposition processing on the AGC frequency modulation command to obtain the initial boundary frequency. The allocation module is used to allocate the AGC frequency modulation command according to the initial boundary frequency to obtain the supercapacitor output command and the lithium battery output command. The correction module includes an acquisition unit, a prediction unit, and a correction unit. The acquisition unit is used to acquire historical charge and discharge current data corresponding to the supercapacitor output command and the lithium battery output command by a bidirectional DC / DC converter circuit. The prediction unit is used to perform temperature change trend prediction processing on the historical charge and discharge current data by combining the temperature data acquired by the thermal management module with a temperature trend prediction algorithm to obtain a temperature trend prediction value. The correction unit is used to perform backtracking correction processing on the initial boundary frequency based on the temperature trend prediction value to obtain a corrected boundary frequency. The processing module is used to allocate and process the AGC frequency modulation command according to the corrected boundary frequency to obtain the supercapacitor target output command and the lithium battery target output command.

[0006] In the technical solution provided in this application, the initial boundary frequency is obtained by performing spectral decomposition processing on the AGC frequency modulation command, and the supercapacitor output command and lithium battery output command are allocated accordingly. This allows the supercapacitor and lithium battery to bear the frequency modulation power of different frequency components according to their respective response speed and energy storage characteristics, avoiding the limitations of a single energy storage form in dealing with the contradictory requirements of both rapid response and continuous output when dealing with the frequency modulation task of a thermal power unit. Furthermore, this application does not stop at completing a one-time allocation according to a fixed boundary frequency, but further uses a bidirectional DC / DC conversion circuit to collect the historical data corresponding to the supercapacitor output command and lithium battery output command. The charging and discharging current data, combined with the temperature data collected by the thermal management module, are used to predict the temperature change trend through a temperature trend prediction algorithm. The predicted temperature trend value is then obtained, and the initial boundary frequency is backtracked to obtain the corrected boundary frequency. The AGC frequency modulation command is then redistributed based on the corrected boundary frequency to obtain the target output command for the supercapacitor and the target output command for the lithium battery. This allows the power allocation result to be dynamically adjusted according to the temperature state of the two types of energy storage units, rather than always being allocated based on a fixed boundary frequency or power boundary. As a result, the supercapacitor and the lithium battery always operate within a power range that matches their current actual state throughout the frequency modulation process.

[0007] In the specific application of the temperature trend prediction algorithm, it is necessary to consider the contribution of the algorithm structure to the overall solution in conjunction with the actual data characteristics of the specific application scenario of frequency regulation of thermal power units. Historical charging and discharging current data and temperature data are correlated over time. The continuous magnitude of the charging and discharging current will gradually accumulate and be transformed into changes in the temperature of the energy storage unit. This accumulation process is not determined by the current or temperature value at a certain moment, but is related to the overall change process of current and temperature within a certain time window. The temperature trend prediction algorithm adopted in this application can capture this ever-changing trend by extracting features and performing time-series modeling on historical charging and discharging current data and temperature data. The cumulative change relationship between the time intervals and the output temperature trend prediction value reflect the direction of temperature change of the energy storage unit in the future, rather than simply recording the current temperature value. This allows for advance processing of the subsequent backoff correction of the boundary frequency, enabling the adjustment of the distribution relationship to be completed before the energy storage unit actually reaches the power boundary. This algorithm feature is consistent with the data pattern of the gradual temperature change of the energy storage unit and the continuous issuance of charging and discharging commands during the frequency regulation of thermal power units. Without considering the time-series accumulation characteristics of the data in this specific application scenario, it would be difficult to achieve the above-mentioned advance adjustment technical effect by simply using instantaneous value judgment. Attached Figure Description

[0008] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a schematic diagram of one embodiment of a supercapacitor and lithium battery hybrid energy storage device for frequency regulation of a thermal power unit in this application. Figure 2 This is a schematic diagram illustrating how the dynamic power boundary of the supercapacitor and lithium battery changes with ambient temperature in an embodiment of this application. Detailed Implementation

[0010] This application provides a hybrid energy storage device combining a supercapacitor and a lithium battery for frequency regulation of a thermal power unit. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0011] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the supercapacitor and lithium battery hybrid energy storage device for frequency regulation of thermal power units in this application includes: The decomposition module is used to perform spectrum decomposition processing on the AGC frequency modulation command to obtain the initial boundary frequency. Specifically, the AGC frequency regulation command is a curve showing the change in active power setpoint of the thermal power unit over time, issued by the power grid dispatching agency. This curve includes both rapid fluctuations caused by random disturbances on the load side and slow changes caused by the unit's own output adjustments. The initial boundary frequency is a frequency value used in this application to distinguish between these two types of components, with a value range set between 0.05 Hz and 0.2 Hz. This range is determined based on statistical observations of historical AGC frequency regulation commands from the thermal power unit. Power fluctuations with periods shorter than 5 seconds account for a significantly higher proportion of the total power fluctuations, belonging to the component dominated by load-side disturbances. Power fluctuations with periods longer than 20 seconds are related to the unit's own adjustment process. Setting the initial boundary frequency in the transition range between the two allows for the differentiation of the load-side disturbance component from the overall command.

[0012] The allocation module is used to allocate the AGC frequency modulation command according to the initial boundary frequency to obtain the supercapacitor output command and the lithium battery output command. Specifically, the supercapacitor output command corresponds to the power component with a frequency higher than the initial boundary frequency, i.e., a power demand with a fast rate of change and a short duration. The lithium battery output command corresponds to the power component with a frequency lower than the initial boundary frequency, i.e., a power demand with a slow rate of change and a long duration. The sum of the two equals the total power value of the AGC frequency modulation command. This constraint relationship ensures that no power deficit or power surplus occurs during the allocation process.

[0013] The correction module includes an acquisition unit, a prediction unit, and a correction unit. The acquisition unit is used to acquire historical charge and discharge current data corresponding to the supercapacitor output command and the lithium battery output command by a bidirectional DC / DC converter circuit. The prediction unit is used to perform temperature change trend prediction processing on the historical charge and discharge current data by combining the temperature data acquired by the thermal management module with a temperature trend prediction algorithm to obtain a temperature trend prediction value. The correction unit is used to perform backtracking correction processing on the initial boundary frequency based on the temperature trend prediction value to obtain a corrected boundary frequency. Specifically, the bidirectional DC / DC converter circuit is a power electronic circuit that connects the supercapacitor, lithium battery and DC bus to realize voltage conversion and bidirectional current flow. The thermal management module is a component arranged in the supercapacitor and lithium battery energy storage compartment to collect the temperature value inside the compartment. The time window length of the historical charge and discharge current data is set to 10 minutes, the sampling interval is set to 1 second, and there are a total of 600 sampling points. The setting of this window length is based on the fact that the thermal time constant of supercapacitor and lithium iron phosphate battery is usually in the range of several minutes to more than ten minutes. If the window is too short, it cannot reflect the heat accumulation process. If the window is too long, it will introduce outdated data. The 10-minute window can cover a typical heat rise process. The temperature trend prediction algorithm consists of three essential layers. The first layer is the feature extraction layer, which performs convolution operations on the sequence of historical charge and discharge current data and temperature data to extract the local correlation features between current and temperature. The second layer is the temporal memory layer, which performs state transfer operations on the feature vector output by the feature extraction layer in chronological order to capture the cumulative trend of temperature changes over time. The third layer is the output layer, which maps the feature vector output by the temporal memory layer to a specific temperature value, i.e., the temperature trend prediction value. This value represents the predicted temperature of the supercapacitor and lithium battery compartment in the future. The specific calculation logic for backtracking the initial boundary frequency based on the temperature trend prediction is as follows: First, the temperature trend prediction is substituted into the power boundary calculation relationship of the supercapacitor and the lithium battery respectively. For every certain degree Celsius increase in temperature, the power boundary value is adjusted down proportionally to obtain the dynamic power boundary after considering the temperature effect. Then, the dynamic power boundary is compared with the power component actually calculated in the allocation process. If the power component exceeds the dynamic power boundary by a larger proportion, the initial boundary frequency will backtrack to a lower frequency direction by a larger amount. The backtracking amount is positively correlated with the excess proportion, thus obtaining the corrected boundary frequency.

[0014] The processing module is used to allocate and process the AGC frequency modulation command according to the corrected boundary frequency to obtain the supercapacitor target output command and the lithium battery target output command.

[0015] Specifically, in the step of allocating the AGC frequency modulation command according to the corrected boundary frequency to obtain the target output command of the supercapacitor and the target output command of the lithium battery, the processing logic is the same as the aforementioned allocation process. The only difference is that the boundary frequency value is replaced by the corrected boundary frequency instead of the initial boundary frequency. The target output command of the supercapacitor and the target output command of the lithium battery are the final power values ​​recalculated after taking into account the temperature change trend. The sum of the two is also equal to the total power value of the AGC frequency modulation command. The power conservation relationship remains unchanged throughout the allocation process.

[0016] In one specific embodiment, the decomposition module is used for: The AGC frequency modulation command is sampled and processed to obtain AGC frequency modulation command time series data; The AGC frequency modulation command time series data is input into a wavelet packet decomposition algorithm for frequency division processing to obtain multiple frequency sub-band power components. The energy percentage of each frequency sub-band is calculated based on the power components of the multiple frequency sub-bands to obtain the energy percentage value of each frequency sub-band. Based on the energy percentage values ​​of each frequency sub-band, a preset energy percentage threshold is compared to obtain a target frequency sub-band that meets the preset energy percentage threshold, and the frequency value corresponding to the target frequency sub-band is determined as the initial boundary frequency.

[0017] Specifically, the decomposition module samples the AGC frequency modulation command. The sampling frequency is set to 10 Hz, which means that a power value is collected every 0.1 seconds. The sampling duration is 60 seconds, resulting in a time series data of the AGC frequency modulation command containing 600 data points. The sampling frequency is set based on the fact that the period of the fastest changing power component in the AGC frequency modulation command is not less than 0.2 seconds. According to the sampling theorem, the sampling frequency must be more than twice the frequency corresponding to this period to avoid signal distortion. The 10 Hz sampling frequency provides sufficient margin. The decomposition module inputs the AGC frequency modulation command time series data into the wavelet packet decomposition algorithm. The decomposition level is set to 3 levels. Each level divides the input signal into two sub-frequency bands according to the frequency and outputs the low-frequency coefficient and high-frequency coefficient respectively. After 3-level decomposition, a total of 8 frequency sub-band power components are generated. The frequency range width corresponding to each sub-band is equal. The reason for setting the decomposition level to 3 levels is that too few levels will result in the sub-band division being too coarse and unable to accurately locate the boundary between the load-side disturbance component and the unit regulation component. Too many levels will result in a significant increase in the amount of computation and the energy proportion of each sub-band being too dispersed, which is not conducive to subsequent threshold comparison.

[0018] The decomposition module accumulates the sum of the squares of the power of each sub-band within the sampling time based on the power components of the eight frequency sub-bands, and divides it by the sum of the squares of the power of all sub-bands to obtain the energy percentage value corresponding to each sub-band. This value represents the proportion of the power fluctuation energy contained in the sub-band in the total AGC frequency modulation command energy. The decomposition module sets a preset energy percentage threshold of 15%. It compares the energy percentage values ​​of the eight sub-bands with this threshold in descending order of frequency. The first sub-band whose energy percentage value exceeds the threshold is identified as the target frequency sub-band. The lower boundary frequency value of the frequency range corresponding to this sub-band is determined as the initial boundary frequency. The rule of comparing in descending order of frequency and selecting the first sub-band that exceeds the threshold ensures that only one target frequency sub-band is identified when the energy percentages of multiple sub-bands are close to the threshold at the same time, avoiding ambiguity in the selection of the boundary frequency. The preset energy percentage threshold of 15% is based on the fact that the sub-band power corresponding to this value usually reflects the critical range where random disturbances on the load side begin to appear in the actual field operation data of thermal power units, but has not yet included the unit's own ramp-up adjustment power.

[0019] In one specific embodiment, the allocation module is used for: Based on the initial boundary frequency, the AGC frequency modulation command is low-pass filtered to obtain the initial output component of the lithium battery. The difference between the AGC frequency modulation command and the initial output component of the lithium battery is calculated to obtain the initial output component of the supercapacitor. Based on the initial output component of the supercapacitor, the current state of charge value of the supercapacitor is subjected to limit judgment processing to obtain the supercapacitor output limit value. The initial output component of the supercapacitor is limited based on the supercapacitor output limit value to obtain the supercapacitor output command. The difference between the initial output component of the lithium battery and the initial output component of the supercapacitor and the supercapacitor output command is superimposed to obtain the lithium battery output command.

[0020] Specifically, the distribution module performs low-pass filtering on the AGC frequency modulation command based on the initial cutoff frequency. The filter used is a second-order Butterworth low-pass filter, with a cutoff frequency equal to the initial cutoff frequency. Power components with frequencies below this cutoff frequency are retained in the output, while those with frequencies above are attenuated. The filtered output signal is the initial power output component of the lithium battery. The second-order Butterworth filter is chosen because it has a narrow attenuation transition band near the cutoff frequency and does not produce significant phase distortion, allowing for a relatively accurate distinction between power components on either side of the cutoff frequency. The distribution module then performs a difference calculation between the AGC frequency modulation command and the initial power output component of the lithium battery. The difference between the power value of the AGC frequency modulation command and the initial power output component of the supercapacitor is the initial power output component. This difference calculation ensures that the sum of the initial power output components of the lithium battery and the supercapacitor is strictly equal to the total power value of the AGC frequency modulation command.

[0021] The distribution module performs over-limit judgment processing on the current state of charge (SOC) value of the supercapacitor based on the initial output component of the supercapacitor. The current SOC value is the percentage of the supercapacitor's remaining usable charge relative to its rated charge, ranging from 0% to 100%. The upper limit of the SOC value is set at 90%, and the lower limit is 10%. The judgment condition is that if the initial output component of the supercapacitor is in the discharge direction and the current SOC value is below 10%, or if the initial output component is in the charging direction and the current SOC value is above 90%, then it is judged as an over-limit. The supercapacitor output limit value output by the over-limit judgment processing is the maximum power value that the supercapacitor is allowed to continue charging and discharging under the current SOC value. When the limit is not exceeded, this value is equal to the rated power of the supercapacitor. When the limit is exceeded, this value decreases proportionally to zero as the SOC value approaches the upper or lower limit. The basis for setting the upper and lower limits of SOC value to 90% and 10% is that deep charging and discharging outside these ranges will accelerate the decomposition rate of the electrolyte inside the supercapacitor and shorten its cycle life. The distribution module limits the initial output component of the supercapacitor based on the supercapacitor output limit value. If the absolute value of the initial output component exceeds the supercapacitor output limit value, the value of the supercapacitor output command is set to the supercapacitor output limit value while retaining the original charging and discharging direction. The difference between the initial output component and the supercapacitor output command is the power that is cut off during the limiting process and cannot be borne by the supercapacitor. The distribution module adds this difference to the initial output component of the lithium battery and transfers this power to the lithium battery to obtain the lithium battery output command. Through this addition process, the sum of the lithium battery output command and the supercapacitor output command is still strictly equal to the total power value of the AGC frequency modulation command, and there will be no power deficit due to the limiting action of the supercapacitor.

[0022] In one specific embodiment, the acquisition unit is used for: The supercapacitor output command is processed by current sampling to obtain a supercapacitor sampling current sequence; The lithium battery output command is processed by current sampling to obtain a lithium battery sampling current sequence; The supercapacitor sampling current sequence and the lithium battery sampling current sequence are time-aligned to obtain a current sampling aligned sequence. The historical charge and discharge current data are obtained by performing sliding truncation processing on the current sampling alignment sequence based on a preset time window.

[0023] Specifically, the acquisition unit performs current sampling processing on the supercapacitor output command. It collects the actual current flowing through the connection branch between the supercapacitor and the bidirectional DC / DC converter circuit at a sampling interval of once per second, obtaining the supercapacitor sampling current sequence. Similarly, the acquisition unit performs current sampling processing on the lithium battery output command. It also collects the actual current flowing through the connection branch between the lithium battery and the bidirectional DC / DC converter circuit at a sampling interval of once per second, obtaining the lithium battery sampling current sequence. Both current sampling processes use the moment the output command is issued as the sampling start time. The sampling interval is set to 1 second because the rate of change of the charging and discharging current of the supercapacitor and lithium battery will not experience drastic jumps within a 1-second timescale. This sampling interval can reflect the current change process without generating too many redundant data points. The acquisition unit performs time alignment processing on the supercapacitor sampling current sequence and the lithium battery sampling current sequence. Since the two current sampling processes are started independently, there is a deviation in the sampling start time due to circuit response delay. The acquisition unit uses the earliest valid sampling time that appears in both sequences as the reference time and removes data points in both sequences that are earlier than the reference time. This ensures that the supercapacitor sampling current sequence and the lithium battery sampling current sequence correspond one-to-one on the time axis, resulting in a current sampling aligned sequence. If this alignment processing is not performed, the subsequent temperature trend prediction algorithm will treat the current values ​​of the two different times as data from the same time when reading the current data, resulting in the prediction result not matching the actual heating situation.

[0024] The acquisition unit performs sliding truncation processing on the current sampling alignment sequence based on a preset time window. The preset time window length is set to 600 seconds, and the sliding step size is set to 60 seconds. That is, every 60 seconds, all data points within the most recent 600 seconds are extracted from the current sampling alignment sequence as one output, resulting in historical charge and discharge current data. The window length is set to 600 seconds because the temperature rise process of supercapacitors and lithium iron phosphate batteries during continuous charge and discharge usually tends to stabilize within a 600-second range. If the window length is too short, the initial heat accumulation process will be missed; if the window length is too long, historical data that has already stabilized will continue to be included in the calculation, increasing the unnecessary computational burden. The sliding step size is set to 60 seconds because this step size allows the historical charge and discharge current data to be continuously updated on a minute-by-minute basis, which matches the calculation cycle required by the subsequent temperature trend prediction algorithm. In the historical charge and discharge current data output after the sliding truncation processing, each data point simultaneously contains the current values ​​of both the supercapacitor and the lithium battery at that moment, for subsequent algorithm reading.

[0025] In one specific embodiment, the prediction unit is used for: The historical charge and discharge current data and the temperature data collected by the thermal management module are spliced ​​together to obtain a temperature-current fusion sequence. The temperature-current fusion sequence is input into the feature extraction layer of the temperature trend prediction algorithm for convolution processing to obtain the temperature-current feature vector. The temperature current feature vector is input into the time-series memory layer of the temperature trend prediction algorithm for time-series modeling processing to obtain the temperature change trend feature vector. The temperature change trend feature vector is input into the output layer of the temperature trend prediction algorithm for regression processing to obtain the temperature trend prediction value.

[0026] Specifically, the prediction unit splices historical charge and discharge current data with temperature data collected by the thermal management module. The thermal management module is placed at one temperature measurement point inside the supercapacitor energy storage compartment and one temperature measurement point inside the lithium battery energy storage compartment. It collects the temperature values ​​inside the compartment at the same 1-second interval as the current sampling, and obtains temperature data that corresponds one-to-one with the time points of the historical charge and discharge current data. After aligning according to the time dimension, the prediction unit arranges the four values ​​of supercapacitor current value, lithium battery current value, supercapacitor compartment temperature value, and lithium battery compartment temperature value together in the feature dimension of the data to form a sequence containing four values ​​at each time point. After arranging 600 time points, a temperature and current fusion sequence with a shape of 600 by 4 is obtained. The prediction unit inputs the temperature-current fusion sequence into the feature extraction layer of the temperature trend prediction algorithm. The feature extraction layer consists of two one-dimensional convolutional layers. The first layer has 16 convolutional kernels with a kernel length of 5 and a stride of 1. After performing convolution operations along the time dimension on the temperature-current fusion sequence, it is connected to a modified linear activation function. The second layer has 32 convolutional kernels with a kernel length of 3 and a stride of 1. It is also connected to a modified linear activation function. The two convolutional operations extract the local change correlation between current and temperature values ​​between adjacent time points layer by layer. After the output of the second convolutional layer, it undergoes one max pooling process with a pooling window length of 2, resulting in a temperature-current feature vector with a length compressed to 150 time points and 32 values ​​per time point.

[0027] The prediction unit inputs the temperature-current feature vector into the temporal memory layer of the temperature trend prediction algorithm. The temporal memory layer consists of a single long short-term memory network with 64 hidden units. The temporal memory layer reads the values ​​of 150 time points from the temperature-current feature vector in chronological order. After reading the value of each time point, it updates the memory state of the current moment by combining it with the memory state passed down from the previous time point. After all 150 time points have been read, the temporal memory layer outputs the 64-dimensional memory state vector corresponding to the last moment, which serves as the temperature change trend feature vector. This vector has already incorporated the cumulative trend information of current and temperature changes within the entire 600-second time window. The prediction unit inputs the temperature change trend feature vector into the output layer of the temperature trend prediction algorithm. The output layer consists of one fully connected layer with an input dimension of 64 and an output dimension of 2, corresponding to the predicted temperature trends of the supercapacitor compartment and the lithium battery compartment, respectively. The fully connected layer does not connect to an activation function, but directly performs a weighted summation operation on the 64-dimensional input vector to output two specific temperature values, namely the temperature trend prediction values. These values ​​represent the prediction results of the temperature values ​​of the supercapacitor compartment and the lithium battery compartment in the future based on the historical data of the current 600 seconds. The training process of the temperature trend prediction algorithm uses historical field operation data as training samples. Each training sample consists of 600 seconds of historical charging and discharging current data and corresponding temperature data as input. The actual measured temperature values ​​of the supercapacitor compartment and lithium battery compartment 300 seconds after the end of the 600-second time window are used as labels. During training, the mean square error is used as the loss function, that is, the average of the squared differences between the two temperature prediction values ​​output by the model and the corresponding label values. The gradient descent method is used to adjust all weight parameters in the convolutional layer, temporal memory layer and output layer to gradually reduce the value of the loss function. The training process continues until the value of the loss function no longer decreases significantly.

[0028] In one specific embodiment, the correction unit is used for: The preset value of the supercapacitor power boundary is corrected based on the temperature trend prediction value to obtain the dynamic power boundary value of the supercapacitor. The preset value of lithium battery power boundary is corrected based on the temperature trend prediction value to obtain the dynamic power boundary value of lithium battery. Based on the dynamic power boundary values ​​of the supercapacitor and the lithium battery, the risk of exceeding the limit is calculated for the initial boundary frequency to obtain the risk value of exceeding the limit. The initial boundary frequency is backed up based on the over-limit risk value to obtain the frequency back-up value. The difference between the initial boundary frequency and the frequency back-up value is calculated to obtain the corrected boundary frequency.

[0029] Specifically, the correction unit corrects the preset value of the supercapacitor power boundary based on the predicted temperature value of the supercapacitor compartment in the temperature trend prediction value. The preset value of the supercapacitor power boundary is 2 megawatts, which is the rated charge and discharge power value of the supercapacitor at an ambient temperature of 25 degrees Celsius. The correction unit calculates the difference between the predicted temperature value of the supercapacitor compartment and 25 degrees Celsius. Whenever the difference increases by 10 degrees Celsius, the preset value of the supercapacitor power boundary is reduced by 5%. The value obtained after correction is the dynamic power boundary value of the supercapacitor. The basis for setting this correction ratio is that the equivalent internal resistance of the electrolyte inside the supercapacitor will increase with the increase of temperature, resulting in a decrease in the current value that can pass under the same voltage conditions. The higher the temperature exceeds 25 degrees Celsius, the more obvious the decrease in usable power. The correction unit adjusts the preset value of the lithium battery power boundary based on the predicted temperature of the lithium battery compartment in the temperature trend prediction value. The preset value of the lithium battery power boundary is 8 megawatts, which is the rated charge and discharge power value of the lithium iron phosphate battery at an ambient temperature of 25 degrees Celsius. The correction unit also calculates the difference between the predicted value of the lithium battery compartment temperature and 25 degrees Celsius. Whenever the difference increases by 10 degrees Celsius, the preset value of the lithium battery power boundary is reduced by 8%. The value obtained after correction is the dynamic power boundary value of the lithium battery. The reason why the reduction ratio of the lithium battery is higher than that of the supercapacitor is that the internal electrochemical reaction rate of the lithium iron phosphate battery is more sensitive to temperature than the physical energy storage process of the supercapacitor. The effect of temperature increase on the reduction of the allowable power value of the lithium battery is more obvious.

[0030] The correction unit performs over-limit risk calculation processing on the initial boundary frequency based on the dynamic power boundary values ​​of the supercapacitor and the lithium battery. The correction unit first re-executes the low-pass filtering and difference calculation processing on the current AGC frequency modulation command according to the initial boundary frequency to obtain the power values ​​that the supercapacitor and the lithium battery should bear at the initial boundary frequency. Then, the power value that the supercapacitor should bear is divided by the supercapacitor dynamic power boundary value, and the power value that the lithium battery should bear is divided by the lithium battery dynamic power boundary value, resulting in two ratios. The larger of these two ratios is taken as the over-limit risk value. An over-limit risk value greater than 1 means that after the temperature changes, the allocation according to the initial boundary frequency will cause the actual power demand of the corresponding energy storage unit to exceed its dynamic power boundary. The correction unit calculates the backoff amount for the initial boundary frequency based on the over-limit risk value. If the over-limit risk value is no greater than 1, the frequency backoff value is 0, and the initial boundary frequency remains unchanged. If the over-limit risk value is greater than 1, the frequency backoff value is equal to the initial boundary frequency multiplied by the portion of the over-limit risk value exceeding 1, multiplied by 0.5. The larger the proportion of the over-limit risk value exceeding 1, the larger the frequency backoff value. The 0.5 proportional coefficient is set to avoid a large one-time change in the frequency backoff value, which would cause drastic fluctuations in the power distribution ratio. The correction unit calculates the difference between the initial boundary frequency and the frequency backoff value, subtracting the frequency backoff value from the initial boundary frequency to obtain the corrected boundary frequency. If the corrected boundary frequency is lower than the initial boundary frequency, it means that the low- and mid-frequency power components originally assigned to the lithium battery will be handled by the supercapacitor. The dynamic power boundary value of the supercapacitor has already considered the actual usable power range after temperature rise, so there will be no situation where the power distribution exceeds the actual capacity of the energy storage unit again.

[0031] Figure 2 This is a schematic diagram illustrating how the dynamic power boundary of the supercapacitor and lithium battery changes with ambient temperature in the embodiments of this application. Figure 2 The figure shows the values ​​of the dynamic power boundary for supercapacitors and lithium batteries in four ambient temperature ranges: 15-25℃, 25-35℃, 35-45℃, and 45-55℃. The preset power boundary values ​​for supercapacitors and lithium batteries are 2 MW and 8 MW, respectively. As the ambient temperature range increases, the dynamic power boundary values ​​for supercapacitors are successively reduced to 1.9 MW, 1.75 MW, and 1.55 MW, while the dynamic power boundary values ​​for lithium batteries are successively reduced to 7.36 MW, 6.56 MW, and 5.6 MW. This figure reflects the correction results of the temperature trend prediction values ​​to the preset power boundary values, providing specific numerical basis for subsequent calculations of over-limit risks and backoff amounts.

[0032] In one specific embodiment, the processing module is configured to: Based on the corrected boundary frequency, the AGC frequency modulation command is low-pass filtered to obtain the corrected output component of the lithium battery. The difference between the AGC frequency modulation command and the lithium battery corrected output component is calculated to obtain the supercapacitor corrected output component. Based on the dynamic power boundary value of the supercapacitor, the over-limit judgment process is performed on the corrected output component of the supercapacitor to obtain the target limit value of the supercapacitor. The supercapacitor's corrected output component is limited based on the supercapacitor's target limiting value to obtain the supercapacitor's target output command. The lithium battery's corrected output component is then superimposed with the difference between the supercapacitor's corrected output component and the supercapacitor's target output command to obtain the lithium battery's target output command.

[0033] Specifically, the processing module performs low-pass filtering on the AGC frequency modulation command based on the corrected cutoff frequency. The filter used is also a second-order Butterworth low-pass filter, but the cutoff frequency is replaced by the corrected cutoff frequency. Power components with frequencies lower than the corrected cutoff frequency are output through the filter, resulting in the corrected output component of the lithium battery. Because the corrected cutoff frequency is lower than the initial cutoff frequency, the frequency range allowed by the filter is correspondingly reduced, and the power component contained in the corrected output component of the lithium battery is less than the initial output component calculated based on the initial cutoff frequency. The processing module performs a difference calculation between the AGC frequency modulation command and the corrected output component of the lithium battery. The power value of the corrected output component of the lithium battery is subtracted from the power value of the AGC frequency modulation command to obtain the corrected output component of the supercapacitor. This difference calculation ensures that the sum of the corrected output component of the lithium battery and the corrected output component of the supercapacitor is still strictly equal to the total power value of the AGC frequency modulation command.

[0034] The processing module performs limit-breaking judgment on the corrected output component of the supercapacitor based on the dynamic power boundary value of the supercapacitor. It compares the absolute value of the corrected output component with the dynamic power boundary value of the supercapacitor. If the absolute value of the corrected output component does not exceed the dynamic power boundary value, the target limit value of the supercapacitor is directly taken as the value of the corrected output component itself. If the absolute value of the corrected output component exceeds the dynamic power boundary value, the target limit value of the supercapacitor is taken as the dynamic power boundary value and the original charging and discharging direction is retained. The judgment is based on the dynamic power boundary value of the supercapacitor rather than a fixed upper and lower limit of the state of charge. The reason is that the dynamic power boundary value of the supercapacitor has already taken into account the impact of temperature changes on the actual usable power range of the supercapacitor in the correction unit. Directly using it can avoid repeated calculations for the same constraint conditions. The processing module limits the supercapacitor's corrected output component based on the supercapacitor's target limiting value. The value of the supercapacitor's target output command is set as the supercapacitor's target limiting value. The difference between the supercapacitor's corrected output component and the supercapacitor's target output command is the portion of power that the supercapacitor cannot handle after the actual usable power of the supercapacitor decreases due to temperature changes. The processing module adds this difference to the lithium battery's corrected output component and transfers this portion of power to the lithium battery to obtain the lithium battery's target output command. The sum of the lithium battery's target output command and the supercapacitor's target output command is still strictly equal to the total power value of the AGC frequency modulation command. The power conservation relationship in the entire distribution process remains consistent before and after the temperature change.

[0035] In one specific embodiment, the target limiting value of the supercapacitor is equal to the dynamic power boundary value of the supercapacitor. The target output command of the supercapacitor and the target output command of the lithium battery are used to control the charging and discharging of the supercapacitor energy storage unit and the lithium iron phosphate battery energy storage unit.

[0036] Specifically, when the absolute value of the corrected output component of the supercapacitor exceeds the dynamic power boundary value of the supercapacitor, the processing module directly sets the target limiting value of the supercapacitor to be equal to the dynamic power boundary value of the supercapacitor, while retaining the original charging and discharging direction of the corrected output component. This equality only holds true in the case of exceeding the limit. If the absolute value of the corrected output component of the supercapacitor does not exceed the dynamic power boundary value of the supercapacitor, the target limiting value of the supercapacitor is still taken as the value of the corrected output component itself, and the two are not equal. After the supercapacitor target output command and the lithium battery target output command are generated, the bidirectional DC / DC converter circuit reads the power value and charging / discharging direction contained in these two output commands respectively, converts them into voltage regulation signals for the corresponding branches, and drives the supercapacitor energy storage unit and the lithium iron phosphate battery energy storage unit to perform charging or discharging actions according to the power value and direction corresponding to their respective target output commands. The bidirectional DC / DC converter circuit determines the current flow direction according to the sign of the output command value. When the value is positive, the corresponding branch performs a discharging action; when the value is negative, the corresponding branch performs a charging action.

[0037] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A hybrid energy storage device combining supercapacitors and lithium batteries for frequency regulation of thermal power units, characterized in that, include: The decomposition module is used to perform spectrum decomposition processing on the AGC frequency modulation command to obtain the initial boundary frequency. The allocation module is used to allocate the AGC frequency modulation command according to the initial boundary frequency to obtain the supercapacitor output command and the lithium battery output command. The correction module includes an acquisition unit, a prediction unit, and a correction unit. The acquisition unit is used to acquire historical charge and discharge current data corresponding to the supercapacitor output command and the lithium battery output command by a bidirectional DC / DC converter circuit. The prediction unit is used to perform temperature change trend prediction processing on the historical charge and discharge current data by combining the temperature data acquired by the thermal management module with a temperature trend prediction algorithm to obtain a temperature trend prediction value. The correction unit is used to perform backtracking correction processing on the initial boundary frequency based on the temperature trend prediction value to obtain a corrected boundary frequency. The processing module is used to allocate and process the AGC frequency modulation command according to the corrected boundary frequency to obtain the supercapacitor target output command and the lithium battery target output command.

2. The supercapacitor and lithium battery hybrid energy storage device for frequency regulation of thermal power units according to claim 1, characterized in that, The decomposition module is used for: The AGC frequency modulation command is sampled and processed to obtain AGC frequency modulation command time series data; The AGC frequency modulation command time series data is input into a wavelet packet decomposition algorithm for frequency division processing to obtain multiple frequency sub-band power components. The energy percentage of each frequency sub-band is calculated based on the power components of the multiple frequency sub-bands to obtain the energy percentage value of each frequency sub-band. Based on the energy percentage values ​​of each frequency sub-band, a preset energy percentage threshold is compared to obtain a target frequency sub-band that meets the preset energy percentage threshold, and the frequency value corresponding to the target frequency sub-band is determined as the initial boundary frequency.

3. The supercapacitor and lithium battery hybrid energy storage device for frequency regulation of thermal power units according to claim 1, characterized in that, The allocation module is used for: Based on the initial boundary frequency, the AGC frequency modulation command is low-pass filtered to obtain the initial output component of the lithium battery. The difference between the AGC frequency modulation command and the initial output component of the lithium battery is calculated to obtain the initial output component of the supercapacitor. Based on the initial output component of the supercapacitor, the current state of charge value of the supercapacitor is subjected to limit judgment processing to obtain the supercapacitor output limit value. The initial output component of the supercapacitor is limited based on the supercapacitor output limit value to obtain the supercapacitor output command. The difference between the initial output component of the lithium battery and the initial output component of the supercapacitor and the supercapacitor output command is superimposed to obtain the lithium battery output command.

4. The supercapacitor and lithium battery hybrid energy storage device for frequency regulation of thermal power units according to claim 1, characterized in that, The acquisition unit is used for: The supercapacitor output command is processed by current sampling to obtain a supercapacitor sampling current sequence; The lithium battery output command is processed by current sampling to obtain a lithium battery sampling current sequence; The supercapacitor sampling current sequence and the lithium battery sampling current sequence are time-aligned to obtain a current sampling aligned sequence. The historical charge and discharge current data are obtained by performing sliding truncation processing on the current sampling alignment sequence based on a preset time window.

5. The supercapacitor and lithium battery hybrid energy storage device for frequency regulation of thermal power units according to claim 4, characterized in that, Prediction unit, used for: The historical charge and discharge current data and the temperature data collected by the thermal management module are spliced ​​together to obtain a temperature-current fusion sequence. The temperature-current fusion sequence is input into the feature extraction layer of the temperature trend prediction algorithm for convolution processing to obtain the temperature-current feature vector. The temperature current feature vector is input into the time-series memory layer of the temperature trend prediction algorithm for time-series modeling processing to obtain the temperature change trend feature vector. The temperature change trend feature vector is input into the output layer of the temperature trend prediction algorithm for regression processing to obtain the temperature trend prediction value.

6. The supercapacitor and lithium battery hybrid energy storage device for frequency regulation of thermal power units according to claim 5, characterized in that, Correction unit, used for: The preset value of the supercapacitor power boundary is corrected based on the temperature trend prediction value to obtain the dynamic power boundary value of the supercapacitor. The preset value of lithium battery power boundary is corrected based on the temperature trend prediction value to obtain the dynamic power boundary value of lithium battery. Based on the dynamic power boundary values ​​of the supercapacitor and the lithium battery, the risk of exceeding the limit is calculated for the initial boundary frequency to obtain the risk value of exceeding the limit. The initial boundary frequency is backed up based on the over-limit risk value to obtain the frequency back-up value. The difference between the initial boundary frequency and the frequency back-up value is calculated to obtain the corrected boundary frequency.

7. The supercapacitor and lithium battery hybrid energy storage device for frequency regulation of thermal power units according to claim 6, characterized in that, Processing module, used for: Based on the corrected boundary frequency, the AGC frequency modulation command is low-pass filtered to obtain the corrected output component of the lithium battery. The difference between the AGC frequency modulation command and the lithium battery corrected output component is calculated to obtain the supercapacitor corrected output component. Based on the dynamic power boundary value of the supercapacitor, the over-limit judgment process is performed on the corrected output component of the supercapacitor to obtain the target limit value of the supercapacitor. The supercapacitor's corrected output component is limited based on the supercapacitor's target limiting value to obtain the supercapacitor's target output command. The lithium battery's corrected output component is then superimposed with the difference between the supercapacitor's corrected output component and the supercapacitor's target output command to obtain the lithium battery's target output command.

8. The supercapacitor and lithium battery hybrid energy storage device for frequency regulation of thermal power units according to claim 7, characterized in that, The target limiting value of the supercapacitor is equal to the dynamic power boundary value of the supercapacitor.

9. The supercapacitor and lithium battery hybrid energy storage device for frequency regulation of thermal power units according to claim 7, characterized in that, The target output command for the supercapacitor and the target output command for the lithium battery are used to control the charging and discharging of the supercapacitor energy storage unit and the lithium iron phosphate battery energy storage unit.

Images