Intelligent control method and system for commercial energy storage

Abstract

The invention discloses an intelligent control method and system for commercial energy storage, and relates to the field of intelligent control, and the method comprises the steps: firstly monitoring the total power consumption of energy consumption equipment, and then carrying out the dynamic adjustment according to the power supply capability of a commercial energy storage system: when the power supply power of the energy storage system is greater than an energy consumption demand, the energy storage system is preferentially used for power supply; when the energy storage electric quantity is lower than a first preset threshold value, switching to a power grid to supply power so as to guarantee continuous operation; when the power supply power of the energy storage system is insufficient to meet the requirement, the energy storage system and the power grid supply power together, and when the energy storage electric quantity is reduced to a second preset threshold value, the power grid charges the energy storage system independently. According to the scheme, the energy utilization efficiency is effectively optimized, the dependence on a power grid is reduced, the stability and economical efficiency of power supply are improved, meanwhile, the service life of the energy storage system is prolonged, and the operation cost is reduced.

Description

Technical Field

[0001] This application relates to the field of intelligent control, and more specifically, to an intelligent control method and system for commercial energy storage. Background Art

[0002] With the popularization of renewable energy and the deepening of power demand side management, the role of commercial energy storage systems in peak shaving and valley filling, and reducing electricity costs has become increasingly prominent. The currently commonly used charging control method is to detect the energy consumption and the connected energy of an enterprise in real time. When the connected energy is greater than the energy consumption, the excess energy is connected to the energy storage device to charge the energy storage device. However, this method has the following problems: When the energy storage device is in a fully charged state, it is unable to store the excess energy, resulting in a large amount of excess energy being consumed, which not only increases the electricity cost but also causes energy waste and reduces the energy utilization efficiency.

[0003] In view of the above technical problems, Chinese Patent CN117498415A proposes a charging control method for an industrial and commercial energy storage system. When the industrial and commercial energy storage system is in a fully charged state and the equipment enters the peak electricity consumption time, according to the comparison result of the power consumption of the equipment and the power supply of the industrial and commercial energy storage system, the industrial and commercial energy storage system is controlled to supply power to the energy-consuming equipment. On the one hand, it can reduce the external energy connected to reduce the electricity cost, and on the other hand, it can improve the energy utilization rate and reduce resource waste.

[0004] However, the above charging control scheme and energy storage control strategy adopt a fixed preset power threshold, which has significant limitations. For example, if the preset power threshold is set too low, the battery will be frequently deeply discharged, which will accelerate its aging and shorten its service life; if the threshold is set too high, it may cause the energy storage system to switch to grid power supply prematurely, unable to make full use of the peak-valley difference in electricity prices, and reducing economic benefits. In addition, the electricity load of commercial users has high dynamics and uncertainty. The fixed preset power value is difficult to adapt to the load fluctuations in different periods and different scenarios, and may intervene in the grid too early or too late in some cases. At the same time, the battery capacity decays with the increase in the number of cycles, and the fixed threshold cannot dynamically match the actual health state of the battery, further exacerbating the risk of control strategy failure.

[0005] Therefore, an optimized intelligent control scheme for commercial energy storage is desired. Summary of the Invention

[0006] To solve the above technical problems, this application is proposed.

[0007] According to one aspect of this application, there is provided an intelligent control method for commercial energy storage, which includes:

[0008] Monitoring and collecting the total power consumption of energy-consuming equipment;

[0009] When the power supply of the commercial energy storage system is greater than the total power consumption, control the commercial energy storage system to supply power to the energy-consuming equipment, and when the real-time power of the commercial energy storage system is lower than the first preset power threshold, use the power grid system to supply power to the energy-consuming equipment and the commercial energy storage system;

[0010] When the power supply of the commercial energy storage system is less than the total power consumption, control the commercial energy storage system and the power grid system to supply power to the energy-consuming equipment, and when the real-time power of the commercial energy storage system is lower than the second preset power threshold, control the power grid system to supply power to the commercial energy storage system;

[0011] Among them, the setting process of the first preset power threshold includes the steps of: extracting the historical power load time series coding features from the time series set of the historical power load data of the energy-consuming equipment; extracting the battery state multi-parameter time series correlation coding features from the time series set of the battery state data of the energy storage system; performing time series collaborative significance feature interaction coding on the historical power load time series coding features and the battery state multi-parameter time series correlation coding features to obtain the power load-battery state time series joint cross-domain dominant interaction coding representation; based on the power load-battery state time series joint cross-domain dominant interaction coding representation, determining the recommended decoding value of the first preset power threshold.

[0012] According to another aspect of the present application, there is provided an intelligent control system for commercial energy storage, which includes:

[0013] A total power consumption acquisition module for monitoring and collecting the total power consumption of the energy-consuming equipment;

[0014] A first power supply module for, when the power supply of the commercial energy storage system is greater than the total power consumption, controlling the commercial energy storage system to supply power to the energy-consuming equipment, and when the real-time power of the commercial energy storage system is lower than the first preset power threshold, using the power grid system to supply power to the energy-consuming equipment and the commercial energy storage system;

[0015] A second power supply module for, when the power supply of the commercial energy storage system is less than the total power consumption, controlling the commercial energy storage system and the power grid system to supply power to the energy-consuming equipment, and when the real-time power of the commercial energy storage system is lower than the second preset power threshold, controlling the power grid system to supply power to the commercial energy storage system;

[0016] Among them, the first power supply module is used for: extracting the historical power load time series coding features from the time series set of the historical power load data of the energy-consuming equipment; extracting the battery state multi-parameter time series correlation coding features from the time series set of the battery state data of the energy storage system; performing time series collaborative significance feature interaction coding on the historical power load time series coding features and the battery state multi-parameter time series correlation coding features to obtain the power load-battery state time series joint cross-domain dominant interaction coding representation; based on the power load-battery state time series joint cross-domain dominant interaction coding representation, determining the recommended decoding value of the first preset power threshold.

[0017] Compared with the prior art, the intelligent control method and system for commercial energy storage provided by the present application first monitor the total power consumption of energy-consuming devices, and then make dynamic adjustments according to the power supply capacity of the commercial energy storage system: when the power supply power of the energy storage system is greater than the energy consumption demand, the energy storage system is preferentially used for power supply, and when the energy storage power is lower than the first preset threshold, the power supply is switched to the power grid to ensure continuous operation; when the power supply power of the energy storage system is insufficient to meet the demand, the energy storage and the power grid jointly supply power, and when the energy storage power drops to the second preset threshold, the power grid charges the energy storage system alone. In the process of setting the above first preset power threshold, by constructing a cross-modal dynamic interaction mechanism between the time-series characteristics of the electrical load and the multi-parameter state of the battery, the spatio-temporal correlation characteristics of the power consumption pattern and the battery state parameters are respectively extracted by using the forward LSTM network and the dilated convolutional neural network. By collaborating on saliency perception and cross-attention calculation, the implicit coupling relationship between the load fluctuation and the battery health state is mined, and finally an adaptive preset power threshold is generated based on dynamic decoding. This method enables the threshold setting to respond to complex working conditions in real time, ensuring the battery life and improving the economic benefits. Description of the Drawings

[0018] By describing the embodiments of the present application in more detail with reference to the accompanying drawings, the above and other objects, features, and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the present application together with the embodiments of the present application, and do not constitute a limitation to the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 It is a flowchart of the intelligent control method for commercial energy storage according to an embodiment of the present application.

[0020] Figure 2 It is a flowchart of the setting process of the first preset power threshold in the intelligent control method for commercial energy storage according to an embodiment of the present application.

[0021] Figure 3 It is a flowchart of step S123 in the intelligent control method for commercial energy storage according to an embodiment of the present application.

[0022] Figure 4 It is a flowchart of step S1233 in the intelligent control method for commercial energy storage according to an embodiment of the present application.

[0023] Figure 5 It is a block diagram of the intelligent control system for commercial energy storage according to an embodiment of the present application. Detailed Description of the Embodiments

[0024] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the accompanying drawings and embodiments of the present disclosure are only for exemplary purposes and are not used to limit the protection scope of the present disclosure.

[0025] It should be understood that the various steps recited in the method embodiments of the present disclosure can be executed in different orders and / or executed in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this regard.

[0026] In the prior art, some studies have attempted to optimize the threshold setting through historical data analysis, but there are still the following problems: The cross-modal correlation between battery state parameters (such as voltage and temperature) and power consumption load has not been fully explored. There is a non-linear complex correlation relationship between the temporal characteristics of the two, resulting in insufficient prediction accuracy and inability to achieve real-time adaptation of the threshold to the system state. Although a few solutions introduce machine learning methods, their model architectures usually lack sufficient consideration of the coupling of temporal feature extraction and cross-modal interaction, restricting the robustness and generalization ability of the control strategy.

[0027] To address the above technical problems, in the technical solution of this application, an intelligent control method for commercial energy storage is proposed. Figure 1 It is a flowchart of an intelligent control method for commercial energy storage according to an embodiment of this application. As Figure 1 shown, the intelligent control method for commercial energy storage according to an embodiment of this application includes: S110, monitoring and collecting the total power consumption of energy-consuming devices; S120, when the power supply of the commercial energy storage system is greater than the total power consumption, controlling the commercial energy storage system to supply power to the energy-consuming devices, and when the real-time power of the commercial energy storage system is lower than the first preset power threshold, using the power grid system to supply power to the energy-consuming devices and the commercial energy storage system; S130, when the power supply of the commercial energy storage system is less than the total power consumption, controlling the commercial energy storage system and the power grid system to supply power to the energy-consuming devices, and when the real-time power of the commercial energy storage system is lower than the second preset power threshold, controlling the power grid system to supply power to the commercial energy storage system.

[0028] In step S110, the total power consumption of the energy-consuming equipment is monitored and collected. In particular, here the detection and collection are carried out when the commercial energy storage system is fully charged and the energy-consuming equipment enters the peak power consumption time. It should be understood that when the commercial energy storage system is in a fully charged state, it is impossible to continue storing excess energy. If effective energy management is not carried out at this time, excess electricity may be wasted, especially during peak power consumption periods, which will lead to low energy utilization efficiency. In addition, peak power consumption times usually mean that the power demand of the energy-consuming equipment reaches its peak, which may require more power supply to meet the demand. By real-time monitoring the total power consumption of the energy-consuming equipment, the current power consumption status can be accurately understood, so as to more effectively allocate the power supply ratio between the energy storage system and the external power grid, and reduce unnecessary external power purchase costs.

[0029] In step S120, when the power supply of the commercial energy storage system is greater than the total power consumption, control the commercial energy storage system to supply power to the energy-consuming equipment, and when the real-time power of the commercial energy storage system is lower than the first preset power threshold, use the power grid system to supply power to the energy-consuming equipment and the commercial energy storage system. Correspondingly, when the power supply capacity of the commercial energy storage system exceeds the current energy consumption demand, preferentially using the power in the energy storage system can effectively consume the stored energy, so as to free up more storage space for emergencies. In order to avoid the inability to respond to emergencies in time (such as power grid failures or emergency power demands) after the commercial energy storage system is completely discharged, a minimum power threshold (i.e., the first preset power threshold) needs to be set. This ensures that even in the event of a power outage, the energy storage system still has a certain amount of reserve energy available for critical equipment.

[0030] The specific implementation process is as follows: First, the control terminal will first detect the power status of the commercial energy storage system. After determining that it is in a fully charged state, it will judge whether the current time node is in the peak power consumption time interval. If the conditions are met, it indicates that the energy-consuming equipment needs to consume a large amount of electric energy, and at this time, excess energy may be generated. However, since the commercial energy storage system is already fully charged, the excess energy cannot be stored. To reduce the call for electric energy from the external power grid and at the same time consume part of the energy storage of the commercial energy storage system for subsequent storage of excess energy, the control terminal will obtain the total power consumption of the energy-consuming equipment in real time. If multiple energy-consuming equipment are running simultaneously, the control terminal will obtain the power consumption of each energy-consuming equipment separately and then sum them up to obtain the total power consumption.

[0031] Next, after obtaining the total power consumption, the control terminal will determine the relationship between the power supply of the commercial energy storage system and the total power consumption. When the power supply of the commercial energy storage system is greater than the total power consumption, it means that the system can power the energy-consuming equipment, thereby consuming part of the energy and freeing up energy storage space. In the process of calling the power of the commercial energy storage system, in order to ensure that the commercial energy storage system can still supply power normally during a power outage and avoid equipment downtime, the control terminal will detect the real-time power of the commercial energy storage system in real time. The first preset power threshold is set based on the interactive analysis between the timing characteristics of the power load and the multi-parameter status of the battery.

[0032] When the real-time power of the commercial energy storage system is lower than the first preset power threshold, the control terminal is connected to the power grid system. The power grid system will supply power to the energy-consuming equipment and the commercial energy storage system respectively, on the one hand to charge the commercial energy storage system, and on the other hand to ensure the normal power consumption of the energy-consuming equipment. If the power connected to the power grid system at this time is greater than the energy consumption of the energy-consuming equipment, the excess power will also be used to charge the commercial energy storage system, thereby improving the utilization efficiency of resources and reducing energy waste.

[0033] In addition, if the commercial energy storage system is not fully charged and the energy-consuming equipment has not entered the peak power consumption period, considering that the cost of non-peak power consumption period is relatively low, the control terminal will control the power grid system to supply power to the commercial energy storage system to fully charge it. After the commercial energy storage system is fully charged, the control terminal will control the power grid system to stop supplying power. When the commercial energy storage system is not fully charged and the energy-consuming equipment enters the peak power consumption period, the control terminal will control the power grid system to supply power to the commercial energy storage system and the energy-consuming equipment at the same time to ensure normal power consumption of the equipment.

[0034] In particular, in the process of setting the above-mentioned first preset power threshold, the technical concept of the present application is to realize the adaptive optimization of the preset power threshold by constructing a cross-modal dynamic interaction mechanism between the time series characteristics of the power load and the multi-parameter state of the battery. Specifically, the deep time series characteristics of the historical power load of the enterprise and the spatiotemporal correlation characteristics of the multi-parameters of the battery state (voltage, temperature) are first integrated, and the nonlinear power consumption pattern is extracted by the forward LSTM network, while the cross-time window correlation of the battery state parameters is captured by the hollow convolutional neural network. Then, by performing collaborative significance perception on the time series characteristics of the historical power load and the time series correlation characteristics of the battery state, the two types of heterogeneous time series features are cross-attention calculated in the time and space dimensions, and the implicit coupling relationship between the load fluctuation trend and the battery health state is mined. Finally, a preset power threshold recommendation value matching the current system state is generated based on the dynamic decoding mechanism. Through the interactive analysis of the multi-modal features of the deep learning model, the threshold setting can respond to complex working conditions such as load mode mutation and battery capacity attenuation in real time, while ensuring the battery life and improving the economic benefits.

[0035] Figure 2It is a flowchart of the setting process of the first preset power threshold in the intelligent control method for commercial energy storage according to an embodiment of the present application. Specifically, in the embodiment of the present application, as Figure 2 shown, the setting process of the first preset power threshold includes the steps: S121, extracting historical power consumption load time series coding features from the time series set of historical power consumption load data of energy-consuming devices; S122, extracting battery state multi-parameter time series correlation coding features from the time series set of battery state data of the energy storage system; S123, performing time series collaborative significance feature interaction coding on the historical power consumption load time series coding features and the battery state multi-parameter time series correlation coding features to obtain a power consumption load-battery state time series joint cross-domain dominant interaction coding representation; S124, based on the power consumption load-battery state time series joint cross-domain dominant interaction coding representation, determining a recommended decoding value of the first preset power threshold.

[0036] Specifically, in step S121, historical power consumption load time series coding features are extracted from the time series set of historical power consumption load data of energy-consuming devices. Specifically, in the embodiment of the present application, extracting historical power consumption load time series coding features from the time series set of historical power consumption load data of energy-consuming devices includes: obtaining the time series set of historical power consumption load data of the energy-consuming device; passing the time series set of the historical power consumption load data through a power consumption load time series feature extractor based on forward LSTM to obtain a historical power consumption load time series coding vector as the historical power consumption load time series coding feature.

[0037] Specifically, in the embodiment of the present application, the time series set of historical power consumption load data of the energy-consuming device is obtained. It should be understood that the time series set of historical power consumption load data of the energy-consuming device refers to a set of power consumption load data of the energy-consuming device recorded in chronological order at different times. These data reflect the change of the power consumption load of the energy-consuming device over time and have the characteristics of a time series. Therefore, by obtaining these data, it is possible to provide original materials for analyzing the power consumption pattern to better capture and excavate the time series feature information of the power consumption load.

[0038] Specifically, various power monitoring devices are usually equipped in commercial premises, such as smart meters, power monitoring systems, etc. These devices will monitor and record the power consumption data of energy-consuming devices in real time. Smart meters can accurately measure the power consumption of devices, and power monitoring systems can not only record power data but also obtain relevant parameters such as power and voltage, which provide reliable original data sources for obtaining historical power consumption load data. In addition, some large commercial buildings or industrial enterprises may also use an energy management system (EMS), which integrates the power consumption data of multiple energy-consuming devices and can be used as an important platform for centralized data acquisition.

[0039] After identifying the data sources, the issue of data collection frequency needs to be addressed. Considering the dynamic change characteristics of commercial users' electricity loads, the collection frequency should be set reasonably. If the collection frequency is too low, important load change information may be missed, and the details of the electricity consumption pattern cannot be accurately captured. If the collection frequency is too high, a large amount of redundant data will be generated, increasing the burden of data storage and processing. Generally speaking, for most commercial energy-consuming devices, a collection frequency at the minute level is more appropriate, such as collecting data every 5 minutes or 10 minutes. This can not only reflect the real-time changes of the load in a timely manner but also not generate too much invalid data.

[0040] Next, connect the data collection terminal to the monitoring device and use a standardized data transmission protocol, such as Modbus, MQTT, etc., to achieve stable data transmission. Taking the Modbus protocol as an example, it is a widely used industrial standard protocol that can conduct data communication between different devices. The data collection terminal sends data request instructions to the monitoring device according to the set collection frequency. After receiving the instructions, the monitoring device returns the corresponding electricity load data to the collection terminal. The collection terminal preliminarily sorts and verifies the collected data to ensure the accuracy and integrity of the data. For example, check whether there are missing values, outliers, etc. in the data. If problems are found, take corresponding processing measures in a timely manner, such as data completion, outlier elimination, etc.

[0041] Store the collected historical electricity load data in a database for subsequent query and analysis. The selection of the database needs to be comprehensively considered according to factors such as the data volume and data processing requirements. For commercial premises with a small data volume, relational databases such as MySQL can be selected, which are simple to operate and easy to maintain. For scenarios with a large data volume and complex processing requirements, time series databases such as InfluxDB can be used. Such databases are specifically optimized for time series data and have high data storage and query performance. When storing data, arrange the data in chronological order and add a timestamp to each piece of data to build a complete time series set of historical electricity load data.

[0042] Finally, after the data acquisition is completed, the data also needs to be updated and maintained regularly. As time goes by, new electricity consumption data is continuously generated, and these data need to be added to the historical dataset in a timely manner to ensure the timeliness of the data. At the same time, regularly check and clean the stored data, delete expired or useless data, release storage space, and improve data processing efficiency.

[0043] Specifically, in the embodiments of the present application, the time series set of the historical power consumption load data is passed through a power consumption load time series feature extractor based on forward LSTM to obtain a historical power consumption load time series encoded vector as the historical power consumption load time series encoded feature. Correspondingly, considering that due to the highly non-linear and dynamically fluctuating characteristics of the power consumption load of commercial users, traditional linear time series models are difficult to effectively capture the implicit periodic, mutational, and trend-related laws. The forward LSTM network can selectively remember long-term dependence features through a gating mechanism, and can extract forward and backward time series patterns from historical load data, such as the distribution of daily load peaks, seasonal power consumption trends, and abnormal fluctuations caused by emergencies. Therefore, in the technical solution of the present application, the time series set of the historical power consumption load data is further passed through a power consumption load time series feature extractor based on forward LSTM to obtain a historical power consumption load time series encoded vector. Through the processing of the power consumption load time series feature extractor based on forward LSTM, the time series set of the original high-dimensional and discrete historical power consumption load data can be transformed into a low-dimensional continuous feature vector, which not only retains the time series context information of the load change, but also implicitly encodes the potential associations between load fluctuations and external factors such as the peak-valley cycle of electricity prices and the operating state of equipment through non-linear mapping.

[0044] Specifically, in step S122, multi-parameter time series correlation encoded features of the battery state are extracted from the time series set of the battery state data of the energy storage system. Specifically, in the embodiments of the present application, extracting multi-parameter time series correlation encoded features of the battery state from the time series set of the battery state data of the energy storage system includes: obtaining the time series set of the battery state data of the energy storage system; performing data arrangement on the time series set of the battery state data based on the time dimension and the sample dimension of the battery state parameters to obtain a multi-parameter time series matrix of the battery state; passing the multi-parameter time series matrix of the battery state through a multi-parameter time series correlation encoder of the battery state based on a dilated convolutional neural network to obtain a multi-parameter time series correlation encoded feature map as the multi-parameter time series correlation encoded features of the battery state.

[0045] Specifically, in the embodiments of the present application, the time series set of the battery state data of the energy storage system is obtained. In particular, the battery state data includes voltage values and temperature values. It should be understood that in a commercial energy storage system, the degradation trajectory of the battery health state is closely related to the dynamic fluctuations and time series changes of the power consumption load. High load periods may accelerate the rise of the battery temperature, while the slow decline trend of the voltage may imply the potential restriction of the battery capacity decay on the power supply capacity. Obtaining these data can provide a basis for studying this complex cross-modal non-linear coupling relationship.

[0046] Specifically, in the embodiments of the present application, the time series set of the battery state data is sorted based on the time dimension and the battery state parameter sample dimension to obtain a battery state multi-parameter time series matrix. Correspondingly, since the time series changes of battery state parameters (such as voltage and temperature) not only have short-period fluctuation characteristics, but also imply long-period degradation patterns caused by battery aging, environmental temperature changes, etc. Traditional time series analysis methods are difficult to simultaneously capture the local mutation characteristics and global decay trends of multiple parameters on the time axis, and independent modeling of a single parameter will ignore the co-variation rules of cross-parameters such as voltage-temperature. Therefore, in the technical solution of the present application, the time series set of the battery state data is sorted based on the time dimension and the battery state parameter sample dimension to obtain a battery state multi-parameter time series matrix. By rearranging the original discretely sampled battery state data into a multi-dimensional matrix according to time steps and parameter types, a spatio-temporal correlation data structure can be constructed, enabling subsequent models to explicitly learn the coupling relationship between different parameters and their dynamic evolution across time scales. The core purpose of performing this data sorting is to provide a spatio-temporal aligned feature expression basis for battery state modeling and eliminate the problem of information fragmentation in the original data caused by sampling frequency differences or parameter heterogeneity.

[0047] Specifically, in the embodiments of the present application, the battery state multi-parameter time series matrix is passed through a battery state multi-parameter time series correlation encoder based on a dilated convolutional neural network to obtain a battery state multi-parameter time series correlation encoded feature map as the battery state multi-parameter time series correlation encoded feature. It should be understood that through the processing of the battery state multi-parameter time series correlation encoder based on the dilated convolutional neural network, the dilation rate adjustment characteristic of the dilated convolution can be utilized to expand the receptive field in the time dimension without increasing the computational amount, effectively capturing the correlation patterns of battery state parameters over a time span of several hours or even several days (such as the slow voltage change caused by the temperature accumulation effect). At the same time, transient events such as voltage dips and temperature anomalies are identified through the local feature extraction ability of multiple convolutional kernels. The generated battery state multi-parameter time series correlation encoded feature map not only fuses the non-linear interaction relationship between multiple parameters of the battery state, but also reveals the potential degradation trajectory of the battery health state and cycle life through cross-time window correlation modeling, providing a battery state representation that dynamically matches the electrical load for subsequent cross-modal interaction, enabling the recommendation of a preset power threshold to synchronously respond to battery capacity decay and external load fluctuations, and avoiding the overcharge or undercharge risks caused by traditional fixed threshold strategies ignoring battery aging.

[0048] Specifically, the step S123 performs time series collaborative significant feature interactive coding on the historical power load time series coding features and the battery state multi-parameter time series associated coding features to obtain the power load-battery state time series joint cross-domain explicit interactive coding representation. Furthermore, considering that there is a complex cross-modal nonlinear coupling relationship between the dynamic fluctuation and time series changes of the power load in the commercial energy storage system and the degradation trajectory of the battery health state, for example, the high load period may accelerate the rise of battery temperature, and the voltage slow drop trend may imply the potential restriction of battery capacity decay on power supply capacity. Due to the lack of effective collaborative modeling of two types of heterogeneous time series data, the existing method cannot capture the dynamic mutual feedback mechanism of load-battery state by the preset threshold. Based on this, in the technical solution of the present application, the historical power load time series coding features and the battery state multi-parameter time series associated coding features are further subjected to time series collaborative significant feature interactive coding to obtain the power load-battery state time series joint cross-domain explicit interactive coding vector as the power load-battery state time series joint cross-domain explicit interactive coding representation. Through the processing of the time series collaborative significant feature interaction encoding, it is possible to utilize a hierarchical feature interaction mechanism (such as asymmetric information enhancement and compression strategy) to refine the local spatiotemporal correlation characteristics of the battery state parameters while retaining the global time series pattern of the power load, and to mine the significant interaction patterns of the two in the key time window through the cross-attention mechanism. For example, the information enhancement of the load feature vector by deconvolution coding enhances its ability to represent the peak and valley cycles of electricity prices and equipment start-stop events, while the compression of the battery state feature map by dilated convolution focuses on the degradation inflection point of the voltage-temperature collaborative change. In terms of execution effect, the generated power load-battery status time series joint cross-domain explicit interaction encoding vector not only captures the long-term dependency between load demand and battery status (such as the impact of high-load operation for consecutive days on battery cycle life) through LSTM recursive modeling, but also filters out redundant noise (such as occasional load spikes or instantaneous temperature fluctuations) through collaborative significance perception in the spatiotemporal dimensions. Ultimately, the dynamic threshold recommendation can synchronously respond to the gradual attenuation of the battery health status and the sudden change in power load demand, avoiding the risk of overcharging or undercharging caused by the traditional fixed threshold strategy due to ignoring cross-modal associations, thereby maximizing economic benefits while extending the battery life.

[0049] Figure 3 Flow chart of step S123 in the intelligent control method for commercial energy storage according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 3As shown in the figure, in step S123, a temporal collaborative significance feature interaction encoding is performed on the historical electricity consumption load temporal coding feature and the battery state multi-parameter temporal correlation coding feature to obtain an electricity consumption load-battery state temporal joint cross-domain dominant interaction coding representation, including: S1231, performing information reduction based on feature separation analysis on the battery state multi-parameter temporal correlation coding feature map to obtain a set of battery state multi-parameter temporal local feature compact coding vectors; S1232, performing information enhancement based on deconvolution coding on the historical electricity consumption load temporal coding vector to obtain a historical electricity consumption load temporal enhanced coding vector; S1233, performing feature layer semantic interaction recursive processing on the set of battery state multi-parameter temporal local feature compact coding vectors and the historical electricity consumption load temporal enhanced coding vector to obtain an electricity consumption load-battery state temporal joint cross-domain dominant interaction coding vector as the electricity consumption load-battery state temporal joint cross-domain dominant interaction coding representation.

[0050] More specifically, in step S1231, performing information reduction based on feature separation analysis on the battery state multi-parameter temporal correlation coding feature map to obtain a set of battery state multi-parameter temporal local feature compact coding vectors includes:

[0051] Performing feature separation on the battery state multi-parameter temporal correlation coding feature map along the channel dimension to obtain a set of battery state multi-parameter temporal local feature coding matrices, which is expressed by the formula:

[0052]

[0053] Wherein, is the battery state multi-parameter temporal correlation coding feature map, is the feature separation process, are respectively the 1st, 2nd, th, and th battery state multi-parameter temporal local feature coding matrices in the set of battery state multi-parameter temporal local feature coding matrices;

[0054] Combining the feature distribution space characteristics of each battery state multi-parameter temporal local feature coding matrix in the set of battery state multi-parameter temporal local feature coding matrices, performing information reduction based on dilated convolution coding on each battery state multi-parameter temporal local feature coding matrix to obtain the set of battery state multi-parameter temporal local feature compact coding vectors, which is expressed by the formula:

[0055] ;

[0056] Wherein, is the A multi-parameter time-series local feature encoding matrix for battery status is the square of the Frobenius norm of the matrix is the dilated convolution encoding is the corresponding compact encoding vector of the multi-parameter time-series local features of the battery status

[0057] It should be understood that the complex coupling relationship among the multi-parameters of the battery status is the key bottleneck hindering the effective extraction of cross-modal features. When directly processing the fused battery status feature map with existing methods, it is often difficult to distinguish the independent dynamic characteristics and their interaction patterns of parameters such as voltage and temperature. For example, the hysteresis effect of the voltage curve during the battery charging and discharging process may show a non-linear correlation with temperature changes, and the coupling of these features in the channel dimension under the traditional encoding method will cause the model to confuse the physical meanings of different parameters. If such cross-parameter correlations cannot be decoupled, subsequent cross-modal interactions will be difficult to accurately capture the true causal relationship between the battery health status and the load demand. For example, the voltage platform shift may be misattributed to temperature fluctuations rather than capacity decay. By decomposing the multi-parameter time-series correlation encoding feature map (including the correlation features of parameters such as voltage and temperature in time series) output by the dilated convolutional neural network into multiple local feature matrices along the channel dimension, each matrix focuses on the spatio-temporal evolution pattern of a specific parameter combination. For example, a certain local matrix may mainly characterize the co-variation of voltage fluctuations and heat dissipation efficiency, while another matrix captures the correlation between temperature gradient and charge-discharge rate. This dissociation operation is essentially a projection of the highly coupled battery status feature space into multiple orthogonal subspaces, enabling the independent analysis of key characteristics such as the time-dependence of voltage parameters and the hysteresis effect of temperature parameters

[0058] Accordingly, battery operating parameters (such as voltage and temperature) have complex spatio-temporal correlation characteristics in the time dimension. If cross-modal interaction is directly performed on the unprocessed local feature matrix, the model will face the dual challenges of high-dimensional data redundancy (such as noise caused by tiny voltage fluctuations in adjacent time steps) and key feature dilution (such as the smoothing of abnormal temperature inflection points), resulting in the subsequent threshold recommendation being unable to accurately capture the early signs of battery performance degradation. For the multi-parameter time-series local feature encoding matrices of each battery state obtained through channel dissociation (such as the sub-matrix focusing on the co-variation of voltage and temperature), according to the spatial characteristics of their feature distributions (such as the range of voltage gradient distribution and the spectral characteristics of temperature fluctuations), the sampling interval and kernel weights of dilated convolution can be dynamically selected. For example, for the voltage feature matrix showing periodic small fluctuations, dilated convolution with a small dilation rate is used to capture the polarization response of adjacent time steps; while for the temperature feature matrix with sudden spikes, convolution with a large dilation rate is used to cross the invalid noise region and focus on the long-term trend of cumulative temperature increase. This adaptive compression mechanism not only eliminates the interference of high-frequency noise on feature expression but also strengthens the cross-time window correlation modeling by expanding the effective receptive field (such as identifying the hidden correlation between the shortening of the voltage plateau period and the intensification of temperature cyclic fluctuations).

[0059] Specifically, in the embodiment of the present application, in step S1232, information enhancement based on deconvolution coding is performed on the historical power consumption load time-series coding vector to obtain a historical power consumption load time-series enhanced coding vector. This process is represented by the formula:

[0060]

[0061] where, is deconvolution coding, and are the trainable deconvolution weight matrix and deconvolution bias vector respectively, is the historical power consumption load time-series coding vector, is the historical power consumption load time-series enhanced coding vector.

[0062] Accordingly, considering that the temporal characteristics of historical electricity load often imply complex dynamic patterns (such as the linkage start and stop of equipment clusters, load transfer driven by electricity price policies), traditional feature extraction methods may lose key temporal details due to dimensionality compression. Transposed convolution coding is not a simple upsampling operation, but rather maps the original encoded vector to a higher-dimensional feature space through learnable convolutional kernel parameters, reactivating the compressed key temporal patterns in the process. For example, for the encoded segment representing the stable operation of low load at night, transposed convolution may enhance its associated features with the charging strategy during the electricity price valley period; while for the load peak segment containing sudden equipment startup, the noise signal and the true load demand features are separated by expanding the feature space. This information enhancement mechanism essentially constructs a dynamic enhancement channel for load features, enabling explicit expression of implicit characteristics such as the timestamp information of equipment start and stop events and the load fluctuation frequency in the feature vector.

[0063] Figure 4 FIG. is a flowchart of step S1233 in the intelligent control method for commercial energy storage according to an embodiment of the present application. More specifically, in the embodiment of the present application, step S1233 performs feature-level semantic interaction recursive processing on the set of battery state multi-parameter temporal local feature compact encoding vectors and the historical electricity load temporal enhanced encoding vectors to obtain an electricity load-battery state temporal joint cross-domain explicit interaction encoding vector, including: S1233-1, performing structured optimization expression on each battery state multi-parameter temporal local feature compact encoding vector in the set of battery state multi-parameter temporal local feature compact encoding vectors to obtain a set of optimized expression battery state multi-parameter temporal local feature compact encoding vectors; S1233-2, respectively inputting the historical electricity load temporal enhanced encoding vector and each optimized expression battery state multi-parameter temporal local feature compact encoding vector in the set of optimized expression battery state multi-parameter temporal local feature compact encoding vectors into a primary interaction module to obtain a set of electricity load-battery state temporal local interaction encoding vectors; S1233-3, inputting the set of electricity load-battery state temporal local interaction encoding vectors into a secondary recursive interaction module based on an LSTM model to obtain the electricity load-battery state temporal joint cross-domain explicit interaction encoding vector.

[0064] More specifically, in the embodiment of the present application, step S1233-1 performs structured optimization expression on each battery state multi-parameter temporal local feature compact encoding vector in the set of battery state multi-parameter temporal local feature compact encoding vectors to obtain a set of optimized expression battery state multi-parameter temporal local feature compact encoding vectors. The specific processing process of this step is as follows:

[0065] If the second-modal local feature encoding matrix to be information-reduced If it is a topological morphological characterization form of the feature space, then the terms for constraint can be regarded as the tight orientation of the overall manifold:

[0066]

[0067] Among them, is the tight orientation constraint factor of the multi-parameter manifold corresponding to the battery state;

[0068] And further based on the multi-parameter topological conserved quantity of the battery state to map the multi-parameter time-series local feature encoding matrix of the battery state from the hybrid topological characterization to the target information-reduced feature mapping, that is to perform the mapping:

[0069]

[0070] Among them, is the multi-parameter topological conserved quantity corresponding to the battery state;

[0071] Thus, the multi-parameter time-series local feature encoding matrix of the battery state has additional topological feature information for the tight orientation constraint factor of the multi-parameter manifold of the battery state and the multi-parameter topological conserved quantity of the battery state so that while retaining the compressed information, it enhances the coupling feature expression of the tight orientation of the manifold and the topological conserved quantity. Therefore, first, for the multi-parameter time-series local feature encoding matrix of the battery state the corresponding multi-parameter standardized information-reduced feature representation vector of the battery state , calculate its multi-parameter external field guiding factor based on the tight orientation constraint factor of the multi-parameter manifold of the battery state :

[0072]

[0073] Among them, is the corresponding multi-parameter standardized information-reduced feature representation vector of the battery state, is the th eigenvalue of the feature vector after function activation of the feature vector , is the multi-parameter external field guiding factor corresponding to the battery state.

[0074] Then, from the multi-parameter topological conserved quantity of the battery state Combined battery state multi-parameter external field guiding factor For more refined representation of higher-dimensional information through the curvature integral characterization method:

[0075] ;

[0076] Among them, is The compressed intermediate coding vector of the time-series local features of the combined battery state multi-parameters after refinement;

[0077] And further optimize the compact coding vector of the time-series local features of the combined battery state multi-parameters for the superposition characterization of the topological subspace as the compressed information representation under the composite topology as:

[0078] ;

[0079] Among them, is The optimized compact coding vector of the time-series local features of the combined battery state multi-parameters with optimized expression;

[0080] In this way, in the topology-enhanced information-reduced representation, the time-series local feature coding matrix of the combined battery state multi-parameters can be better expressed The structural information of, thereby improving the information-reduced expression effect of the second-modal local feature compressed coding vector.

[0081] More specifically, in the embodiment of the present application, in step S1233-2, each optimized compact coding vector of the time-series local features of the combined battery state multi-parameters in the set of the historical power consumption load time-series enhanced coding vector and the optimized compact coding vector of the time-series local features of the combined battery state multi-parameters is respectively input into the primary interaction module to obtain a set of power consumption load-battery state time-series local interaction coding vectors, and this process is expressed by the formula as:

[0082]

[0083]

[0084] Among them, is feature concatenation, is the th optimized compact coding vector of the time-series local features of the combined battery state multi-parameters in the set of optimized compact coding vectors of the time-series local features of the combined battery state multi-parameters, is dot product by position, is subtraction by position, and are the trainable interaction weight matrix and the trainable interaction bias vector respectively, is the Sigmoid activation function, the 1st, 2nd, rd, and th electrical load-battery state time series local interaction coding vectors respectively in the set of electrical load-battery state time series local interaction coding vectors, and the set is the set of the electrical load-battery state time series local interaction coding vectors.

[0085] It should be understood that the heterogeneity of load and battery state characteristics and the complexity of local correlation are the key bottlenecks restricting the dynamic optimization of thresholds. When existing methods directly interact the global load characteristics with the overall battery state characteristics, it is difficult to capture the coupling relationship in different parameter subspaces under a specific time window (such as the local correlation between load peak periods and sudden increases in battery temperature), resulting in the interaction result tending to the global average effect and ignoring key details. Without a fine-grained local interaction mechanism, the model may mistakenly regard the accidental synchronous fluctuations of load-battery state as regular correlations, or ignore the impact of small voltage platform offsets on threshold setting in high-frequency shallow charge and discharge scenarios. The enhanced coding vector of the historical electrical load time series after deconvolution excitation carries the enhanced global time series pattern (such as a continuous high-load operation cycle), while the set of compact coding vectors of the optimized expression of the multi-parameter time series local characteristics of the battery state after structural optimization represents the refined state evolution of different parameter subspaces (such as the specific phase relationship between voltage-temperature co-variation). The primary interaction unit independently calculates each group of load-battery feature pairs. For example, it assigns attention weights to the coding segment representing the basic load at night and the self-discharge characteristics of the battery in the static state, or scores the correlation between the load sudden increase segment at noon and the temperature gradient change caused by the large current discharge of the battery. This parallel processing mechanism enables the model to identify diverse local mapping relationships between load demands and battery states (such as preferentially associating with voltage stability characteristics during the load slow increase stage and paying attention to temperature recovery characteristics during the load sudden drop stage), avoiding pattern confusion caused by the feature drowning effect in global interaction.

[0086] More specifically, in the embodiment of the present application, the set of the electrical load-battery state time series local interaction coding vectors is input into the secondary recursive interaction module based on the LSTM model to obtain the electrical load-battery state time series joint cross-domain explicit interaction coding vector, and this process is represented by the formula:

[0087]

[0088] wherein, is processed by the model, and is the electrical load-battery state time series joint cross-domain explicit interaction coding vector.

[0089] Finally, considering that although the primary interaction module can capture the local correlations between the load and battery state within a specific time window (such as the correspondence between a sudden increase in load and a sharp rise in battery temperature), it cannot model the dynamic cumulative effects across time periods (such as the impact of capacity decay caused by consecutive days of shallow charge and discharge on subsequent threshold settings). When processing a sequence of local interaction coding vectors, the LSTM network uses its gating mechanism (forget gate, input gate, output gate) to dynamically screen key interaction patterns and establish long-term temporal dependencies. For example, when processing the local interaction vector corresponding to a deep discharge event, the LSTM can combine the frequency records of similar events in historical interactions (such as the number of deep discharges in the past 30 days) to evaluate the cumulative impact of the current discharge behavior on battery life. At the same time, for occasional load spikes and abnormal battery parameters (such as transient temperature overlimit), the long-short-term memory ability of the LSTM can effectively suppress the interference of such noise on global decisions. This recursive fusion mechanism enables the model to identify the key evolution paths in the load-battery state interaction (such as the implicit correlation between the slope change of the load demand curve and the battery capacity decay rate), thereby constructing a temporally coherent cross-modal semantic representation.

[0090] Specifically, in step S124, based on the time-series joint cross-domain explicit interaction coding representation of the electricity load-battery state, a recommended decoding value of the first preset power threshold is determined. Specifically, in the embodiments of the present application, determining a recommended decoding value of the first preset power threshold based on the time-series joint cross-domain explicit interaction coding representation of the electricity load-battery state includes: passing the time-series joint cross-domain explicit interaction coding vector of the electricity load-battery state through a preset power recommender based on a decoder to obtain the recommended decoding value of the first preset power threshold. It should be understood that although the time-series joint cross-domain explicit interaction coding vector of the electricity load-battery state integrates the spatio-temporal correlation features of load fluctuations and battery states, its essence is still a high-dimensional implicit representation and cannot be directly mapped to an operable preset power threshold. Existing threshold setting methods often rely on empirical formulas or static rules and cannot transform complex cross-modal interaction patterns (such as the instantaneous capacity decay caused by the cumulative effect of battery temperature during a load surge period) into dynamic threshold adjustment strategies. Therefore, in the technical solution of the present application, the time-series joint cross-domain explicit interaction coding vector of the electricity load-battery state is further passed through a preset power recommender based on a decoder to obtain the recommended decoding value of the first preset power threshold. Through the preset power recommender based on a decoder, structures such as deconvolution or fully connected layers can be used to decode the cross-modal association rules (such as the constraint relationship between the battery health state and the depth of discharge during peak load periods) implicit in the time-series joint cross-domain explicit interaction coding vector of the electricity load-battery state into threshold recommendation values matching the current system state.

[0091] In summary, the setting process of the first preset power threshold is explained, which is to achieve adaptive optimization of the preset power threshold by constructing a cross-modal dynamic interaction mechanism between the time series characteristics of power load and the multi-parameter state of the battery. Specifically, the deep time series characteristics of the historical power load of the enterprise and the spatiotemporal correlation characteristics of the multi-parameters of the battery state (voltage, temperature) are first integrated, and the nonlinear power consumption pattern is extracted by the forward LSTM network, while the cross-time window correlation of the battery state parameters is captured by the hollow convolutional neural network. Then, by performing collaborative significance perception on the time series characteristics of the historical power load and the time series correlation characteristics of the battery state, the two types of heterogeneous time series features are cross-attention calculated in the time and space dimensions, and the implicit coupling relationship between the load fluctuation trend and the battery health state is mined. Finally, the preset power threshold recommendation value matching the current system state is generated based on the dynamic decoding mechanism. Through the interactive analysis of the multi-modal features of the deep learning model, the threshold setting can respond to complex working conditions such as load mode mutation and battery capacity attenuation in real time, while ensuring the battery life and improving the economic benefits.

[0092] In step S130, when the power supply of the commercial energy storage system is less than the total power consumption, the commercial energy storage system and the power grid system are controlled to supply power to the energy-consuming equipment, and when the real-time power of the commercial energy storage system is lower than the second preset power threshold, the power grid system is controlled to supply power to the commercial energy storage system. In particular, it is worth mentioning that the setting process of the second preset power threshold is similar to the setting process of the first preset power threshold. It should be understood that when the power supply of the commercial energy storage system is less than the total power consumption of the energy-consuming equipment, in this case, the energy storage system alone cannot meet the needs of all energy-consuming equipment. In order to avoid the complete discharge of the energy storage system and affect its use in emergency situations, it is necessary to set a minimum power threshold (i.e., the second preset power threshold) to ensure that the system always has a certain amount of reserve power. In this way, timely access to the power grid for charging prevents the energy storage system from being too low in power, helps to maintain its good working condition, prolongs its service life, and ensures that there is sufficient capacity in the future to cope with peak demand or emergencies.

[0093] The specific implementation is as follows: First, when the commercial energy storage system is fully charged and the energy-consuming equipment enters the peak power consumption period, the control terminal will obtain the total power consumption of the energy-consuming equipment in real time. Taking a large commercial center as an example, it contains many shops, lighting equipment, air conditioning systems and other energy-consuming equipment. The control terminal will collect and integrate the power consumption data of these devices to obtain the total power consumption.

[0094] Next, after determining that the power supply of the commercial energy storage system is less than the total power consumption, the electricity consumption type is determined as commercial electricity. At this time, the historical electricity consumption power of the commercial building is obtained from the preset database. For example, by obtaining the address information of the commercial center, the corresponding electricity bill table is searched in the database, the electricity bill amount information is extracted from the table, the historical electricity consumption is calculated, and then the historical electricity consumption power is obtained through conversion. Assume that the electricity consumption power on a certain day in the same period of the previous year of this commercial center is the historical electricity consumption power data.

[0095] Then, the power difference is calculated using the historical electricity consumption power and the power supply of the commercial energy storage system, and then the ratio of the power difference to the power supply is calculated to obtain the power supply ratio. For example, if the power supply of the commercial energy storage system is 800 kW and the historical electricity consumption power is 600 kW, the power difference is 200 kW, and the calculated power supply ratio is 200÷800 = 25%. If it is found that the power difference is large, the power supply ratio will be adjusted appropriately according to the actual situation to avoid excessive power consumption of the commercial energy storage system.

[0096] After determining the power supply ratio, the control terminal controls the commercial energy storage system and the power grid system to supply power to the energy-consuming equipment according to this ratio. During this process, the real-time power of the commercial energy storage system is continuously monitored. The setting process of the second preset power threshold is similar to that of the first preset power threshold, which is used to ensure that the equipment can still operate normally for a period of time in case of sudden power failure.

[0097] When the real-time power of the commercial energy storage system is lower than the second preset power threshold, the control terminal will immediately control the power grid system to supply power to the commercial energy storage system. For example, when it is monitored that the power of the commercial energy storage system drops to be close to or lower than the second preset power threshold, the power grid system starts the charging operation of the commercial energy storage system, and at the same time continues to ensure the power supply of the energy-consuming equipment. During the charging process of the commercial energy storage system, the control terminal continuously monitors its power change until it is fully charged.

[0098] In addition, if the commercial energy storage system is in a state of not being fully charged and the energy-consuming equipment has not entered the peak electricity consumption time, the control terminal will control the power grid system to supply power to the commercial energy storage system to charge it to the full charge state. When the commercial energy storage system is fully charged, the control terminal controls the power grid system to stop supplying power.

[0099] In summary, the intelligent control method for commercial energy storage according to the embodiments of the present application is elucidated. It first monitors the total power consumption of energy-consuming devices and then makes dynamic adjustments according to the power supply capacity of the commercial energy storage system: when the power supply of the energy storage system is greater than the energy consumption demand, the energy storage system is preferentially used for power supply, and when the energy storage power is lower than the first preset threshold, it switches to grid power supply to ensure continuous operation; when the power supply of the energy storage system is insufficient to meet the demand, the energy storage and the grid supply power together, and when the energy storage power drops to the second preset threshold, the grid charges the energy storage system alone. This solution effectively optimizes the energy utilization efficiency, reduces the dependence on the grid, improves the stability and economy of power supply, extends the service life of the energy storage system, and reduces the operating cost.

[0100] Figure 5 FIG. is a block diagram of an intelligent control system for commercial energy storage according to an embodiment of the present application. As Figure 5 shown, the intelligent control system 100 for commercial energy storage according to an embodiment of the present application includes: a total power consumption acquisition module 110 for monitoring and collecting the total power consumption of energy-consuming devices; a first power supply module 120 for controlling the commercial energy storage system to supply power to the energy-consuming devices when the power supply of the commercial energy storage system is greater than the total power consumption, and using the grid system to supply power to the energy-consuming devices and the commercial energy storage system when the real-time power of the commercial energy storage system is lower than the first preset power threshold; a second power supply module 130 for controlling the commercial energy storage system and the grid system to supply power to the energy-consuming devices when the power supply of the commercial energy storage system is less than the total power consumption, and controlling the grid system to supply power to the commercial energy storage system when the real-time power of the commercial energy storage system is lower than the second preset power threshold.

[0101] Here, those skilled in the art can understand that the specific operations of each step in the above intelligent control system for commercial energy storage have been described in detail in the description of the intelligent control method for commercial energy storage above with reference to Figures 1 to 4 and thus, the repeated description thereof will be omitted.

[0102] As described above, the intelligent control system 100 for commercial energy storage according to the embodiments of the present disclosure can be implemented in various wireless terminals, such as a server with an intelligent control algorithm for commercial energy storage. In a possible implementation, the intelligent control system 100 for commercial energy storage according to the embodiments of the present disclosure can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the intelligent control system 100 for commercial energy storage can be a software module in the operating system of the wireless terminal, or can be an application program developed for the wireless terminal; of course, the intelligent control system 100 for commercial energy storage can also be one of the many hardware modules of the wireless terminal.

[0103] Alternatively, in another example, the intelligent control system 100 for commercial energy storage and the wireless terminal may also be separate devices, and the intelligent control system 100 for commercial energy storage can be connected to the wireless terminal through a wired and / or wireless network and transmit interaction information according to a predefined data format.

[0104] In summary, it is intended that the above detailed description be considered illustrative rather than restrictive, and it should be understood that the above embodiments should be construed as merely illustrative of the present invention and not as limiting the scope of the present invention. After reading the content described in the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. An intelligent control method for commercial energy storage, characterized in that Including: Monitoring and collecting the total power consumption of energy-consuming devices; When the power supply of the commercial energy storage system is greater than the total power consumption, controlling the commercial energy storage system to supply power to the energy-consuming devices, and when the real-time power of the commercial energy storage system is lower than the first preset power threshold, using the power grid system to supply power to the energy-consuming devices and the commercial energy storage system; When the power supply of the commercial energy storage system is less than the total power consumption, controlling the commercial energy storage system and the power grid system to supply power to the energy-consuming devices, and when the real-time power of the commercial energy storage system is lower than the second preset power threshold, controlling the power grid system to supply power to the commercial energy storage system; Among them, the setting process of the first preset power threshold includes the steps of: extracting the historical power load time series coding features from the time series set of the historical power load data of the energy-consuming devices; extracting the battery state multi-parameter time series correlation coding features from the time series set of the battery state data of the energy storage system; performing time series collaborative significance feature interaction coding on the historical power load time series coding features and the battery state multi-parameter time series correlation coding features to obtain the power load-battery state time series joint cross-domain dominant interaction coding representation; based on the power load-battery state time series joint cross-domain dominant interaction coding representation, determining the recommended decoding value of the first preset power threshold.

2. The intelligent control method for commercial energy storage according to claim 1, wherein The battery state data includes voltage value and temperature value.

3. The intelligent control method for commercial energy storage according to claim 2, characterized in that Extracting the historical power load time series coding features from the time series set of the historical power load data of the energy-consuming devices includes: Obtaining the time series set of the historical power load data of the energy-consuming devices; Passing the time series set of the historical power load data through a power load time series feature extractor based on forward LSTM to obtain the historical power load time series coding vector as the historical power load time series coding features.

4. The intelligent control method for commercial energy storage according to claim 3, wherein Extracting the battery state multi-parameter time series correlation coding features from the time series set of the battery state data of the energy storage system includes: Obtaining the time series set of the battery state data of the energy storage system; Sorting out the time series set of the battery state data based on the time dimension and the battery state parameter sample dimension to obtain the battery state multi-parameter time series matrix; Passing the battery state multi-parameter time series matrix through a battery state multi-parameter time series correlation encoder based on a dilated convolutional neural network to obtain the battery state multi-parameter time series correlation coding feature map as the battery state multi-parameter time series correlation coding features.

5. The intelligent control method for commercial energy storage according to claim 4, wherein Performing time series collaborative significance feature interaction coding on the historical power load time series coding features and the battery state multi-parameter time series correlation coding features to obtain the power load-battery state time series joint cross-domain dominant interaction coding representation includes: Performing information reduction based on feature separation analysis on the battery state multi-parameter time series correlation coding feature map to obtain a set of battery state multi-parameter time series local feature compact coding vectors; Performing information enhancement based on transposed convolution coding on the historical power load time series coding vector to obtain the historical power load time series enhanced coding vector; Perform feature-level semantic interaction recursive processing on the set of compact encoded vectors of multi-parameter time-series local features of the battery state and the enhanced encoded vector of the historical power consumption load time series to obtain a power consumption load-battery state time-series joint cross-domain explicit interaction encoded vector as the power consumption load-battery state time-series joint cross-domain explicit interaction encoding representation.

6. The intelligent control method for commercial energy storage according to claim 5, characterized in that, Perform information reduction based on feature separation analysis on the multi-parameter time-series correlation encoded feature map of the battery state to obtain a set of compact encoded vectors of multi-parameter time-series local features of the battery state, including: Perform feature separation on the multi-parameter time-series correlation encoded feature map of the battery state along the channel dimension to obtain a set of multi-parameter time-series local feature encoded matrices of the battery state; Combining the characteristic distribution space characteristics of each multi-parameter time-series local feature encoded matrix in the set of multi-parameter time-series local feature encoded matrices of the battery state, perform information reduction on each multi-parameter time-series local feature encoded matrix based on dilated convolution encoding to obtain the set of compact encoded vectors of multi-parameter time-series local features of the battery state.

7. The intelligent control method for commercial energy storage according to claim 6, wherein, Perform feature-level semantic interaction recursive processing on the set of compact encoded vectors of multi-parameter time-series local features of the battery state and the enhanced encoded vector of the historical power consumption load time series to obtain a power consumption load-battery state time-series joint cross-domain explicit interaction encoded vector, including: Perform structured optimization expression on each compact encoded vector of multi-parameter time-series local features of the battery state in the set of compact encoded vectors of multi-parameter time-series local features of the battery state to obtain a set of optimized expression compact encoded vectors of multi-parameter time-series local features of the battery state; Input the enhanced encoded vector of the historical power consumption load time series and each optimized expression compact encoded vector of multi-parameter time-series local features of the battery state in the set of optimized expression compact encoded vectors of multi-parameter time-series local features of the battery state into the primary interaction module respectively to obtain a set of power consumption load-battery state time-series local interaction encoded vectors; Input the set of power consumption load-battery state time-series local interaction encoded vectors into the secondary recursive interaction module based on the LSTM model to obtain the power consumption load-battery state time-series joint cross-domain explicit interaction encoded vector.

8. The intelligent control method for commercial energy storage according to claim 7, characterized in that, Based on the power consumption load-battery state time-series joint cross-domain explicit interaction encoding representation, determine the recommended decoding value of the first preset power threshold, including: passing the power consumption load-battery state time-series joint cross-domain explicit interaction encoded vector through a preset power threshold recommender based on a decoder to obtain the recommended decoding value of the first preset power threshold.

9. An intelligent control system for commercial energy storage, characterized in that, Including: A total power consumption acquisition module for monitoring and collecting the total power consumption of energy-consuming devices; A first power supply module for controlling the commercial energy storage system to supply power to the energy-consuming devices when the power supply power of the commercial energy storage system is greater than the total power consumption, and using the power grid system to supply power to the energy-consuming devices and the commercial energy storage system when the real-time power of the commercial energy storage system is lower than the first preset power threshold. The second power supply module is used to control the commercial energy storage system and the power grid system to supply power to the energy-consuming equipment when the power supply power of the commercial energy storage system is less than the total power consumption, and to control the power grid system to supply power to the commercial energy storage system when the real-time power of the commercial energy storage system is lower than the second preset power threshold; Among them, the first power supply module is used to: extract the historical power load time series coding features from the time series set of the historical power load data of the energy-consuming equipment; extract the battery state multi-parameter time series correlation coding features from the time series set of the battery state data of the energy storage system; perform time series collaborative significance feature interaction coding on the historical power load time series coding features and the battery state multi-parameter time series correlation coding features to obtain the power load-battery state time series joint cross-domain dominant interaction coding representation; based on the power load-battery state time series joint cross-domain dominant interaction coding representation, determine the recommended decoding value of the first preset power threshold.

Images