Energy storage power station management system based on EMS

Abstract

The invention, which relates to the technical field of energy storage power station management, discloses an EMS-based energy storage power station management system comprising the following components: a data acquisition module, a multi-dimensional space-time sequence prediction module, an energy storage scheduling strategy making module and an execution and monitoring module. Comprehensive time dimension and space dimension data are collected through the data acquisition module, accurate prediction is performed through the multi-dimensional space-time sequence prediction module, the charging and discharging requirements of the energy storage power station can be accurately predicted, the energy storage scheduling strategy making module makes an optimal scheduling strategy according to a prediction result, efficient operation of the energy storage power station is achieved, and the energy storage power station is ensured to be stable and reliable. The capacity of the energy storage power station can be utilized to the maximum extent, the charging and discharging efficiency is improved, discharging and charging plans can be made according to the electricity prices in different time periods, the operation cost is reduced, and economic benefits are increased.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage power station management, and particularly to an energy storage power station management system based on EMS. Background Art

[0002] With the large-scale grid connection of renewable energy and the rapid development of distributed energy of electric vehicles, as an important part of the power grid, the management and dispatching of energy storage power stations have become increasingly complex. Energy storage power stations not only need to optimize charge and discharge strategies in the time and space dimensions, but also need to consider various factors such as interaction with the power grid and weather changes.

[0003] Traditional technologies have disadvantages. On the one hand, these methods often ignore the mutual influence between time and geographical location, resulting in inaccurate prediction results. On the other hand, traditional methods lack flexibility in processing different types of data and are difficult to adapt to the diversity of data characteristics. In addition, in the process of formulating traditional energy storage dispatching strategies, there is often a lack of real-time and intelligence, making it difficult to respond in a timely manner to changes in power grid load and actual demands of energy storage power stations, thus affecting the operation efficiency and economic benefits of energy storage power stations.

[0004] In summary, traditional technologies have many limitations in processing energy storage power station data and are difficult to meet the requirements of modern power grids for efficient and intelligent management of energy storage power stations. Therefore, it is particularly important to develop an energy storage power station management system based on EMS. Summary of the Invention

[0005] The purpose of the present invention is to make up for the deficiencies of the existing technology and provide an energy storage power station management system based on EMS. It can achieve accurate prediction, efficient dispatching and real-time monitoring of energy storage power stations through data acquisition, multi-dimensional spatio-temporal sequence prediction, energy storage dispatching strategy formulation and execution and monitoring modules, which is of great significance for improving the operation efficiency of energy storage power stations, reducing operation costs and enhancing power grid stability.

[0006] To solve the above technical problems, the present invention provides the following technical solution: An energy storage power station management system based on EMS, which includes the following components: a data acquisition module, a multi-dimensional spatio-temporal sequence prediction module, an energy storage dispatching strategy formulation module, and an execution and monitoring module;

[0007] The data acquisition module: collects historical power data in the time dimension, electricity price fluctuation data, weather difference data at different geographical locations in the space dimension, and surrounding power grid load distribution data, and the data is collected in real time through sensors, weather stations, and power grid monitoring devices distributed at different positions;

[0008] The multi-dimensional spatio-temporal sequence prediction module: preprocesses the collected data, unifies different types of data under the same dimension, and sets the preprocessed data vector as Xt,s = [P t , E t , W s , G s , construct a multi-dimensional spatio-temporal sequence prediction model, and introduce a new spatio-temporal interaction factor θ t,s , which reflects the degree of mutual influence between time t and geographical location s. The prediction model is based on the following formula:

[0009]

[0010] where, is the predicted value of the charge and discharge demand of the energy storage power station at time t + 1 and geographical location s. n and m are the historical data backtracking steps in the time and space dimensions respectively, and ω i,j is the weight coefficient, indicating the influence degree of historical data X t-i,s-j on the predicted value. f(·) is a non-linear transformation function, and is used, where k is adaptively adjusted according to the battery aging degree;

[0011] The spatio-temporal interaction factor θ t,s is calculated as follows:

[0012]

[0013] where, q is the number of factors affecting spatio-temporal interaction, α k is the weight of each factor, and δ k (t, s is the influence index of the kth factor at time t and geographical location s. When k = 1 represents the interaction influence between weather and power, δ 1 (t, s) can be obtained through the correlation calculation of light intensity L s and historical power P t ;

[0014] The weight coefficients ω i,j and α k are determined by an adaptive weight adjustment algorithm, which is iteratively adjusted based on the feedback of the prediction error. Let the prediction error at the lth iteration be Y t+1,s be the actual charge and discharge demand, then the weight update formula is:

[0015]

[0016] where, β and γ are the learning rates, which are determined through experiments and experience, and control the adjustment speeds of weights ω i,j and α k respectively;

[0017] The energy storage scheduling strategy formulation module: based on the charge and discharge demand prediction results output by the multi-dimensional spatiotemporal sequence prediction module Combined with the current battery power state SOC of the energy storage power station t , Battery charge and discharge power limit P max and P min , the operating health status H of the equipment, formulate the energy storage scheduling strategy, and use a comprehensive optimization objective function F to determine the optimal scheduling strategy:

[0018] F=λ 1 ·(RC)+λ 2 ·(SOC t+1 -SOC optimal ) 2 +λ 3 ·H

[0019] Among them, R is the discharge benefit, C is the charging cost, SOC optimal is the optimal state of charge of the battery, 1 , 2 and λ 3 is the weight coefficient, which is determined by evaluating the importance of different operating objectives of the energy storage power station. The optimization objective function is solved using the gradient descent method to obtain the optimal charging and discharging time and power scheduling parameters;

[0020] The execution and monitoring module is responsible for executing the scheduling strategy generated by the energy storage scheduling strategy formulation module, controlling the operation of the charging and discharging equipment of the energy storage power station, and at the same time, monitoring the operating status of the energy storage power station in real time, and feeding back these data to the data acquisition module and the multi-dimensional space-time series prediction module, so as to adjust and optimize the prediction model and scheduling strategy in real time.

[0021] Furthermore, the data acquisition module, for the historical power data P t The multi-resolution sampling strategy is used to collect data during peak power consumption periods at a higher sampling frequency to capture rapid changes in power. During low power consumption periods, a lower sampling frequency is used to collect data on electricity price fluctuations. t , and also collects historical electricity price adjustment policies and trend information of the electricity market to build a background knowledge base for electricity price forecasting. s In terms of collection, the humidity H s , air pressure P s Meteorological parameter collection, surrounding power grid load distribution data G s By exchanging data with multiple power grid dispatch centers at different levels, more comprehensive load information can be obtained. At the same time, a preliminary quality assessment of the collected data is conducted to eliminate obviously erroneous or abnormal data, providing an accurate and reliable data basis for subsequent prediction and dispatch.

[0022] Furthermore, in the data preprocessing part of the multi-dimensional spatio-temporal sequence prediction module, a preprocessing method based on data feature clustering is adopted. First, feature extraction is performed on different types of collected data. For power data P t , its mean, variance, and peak statistical features are extracted. For weather data W s , the correlation features between different meteorological parameters are extracted. Then, the data is clustered according to these features into different categories. For each category, different normalization methods are used for processing. For the data category with large fluctuations, power data, an adaptive normalization method is adopted, and the normalization scale is dynamically adjusted according to the real-time fluctuation range of the data. For the relatively stable data category, some meteorological data adopts the traditional linear normalization method. In addition, during the preprocessing process, the time series characteristics of the data are also considered, and missing data is interpolated and filled using an interpolation method based on spatio-temporal correlation, that is, the missing values are estimated according to the data correlation at different times in the same geographical location and at different geographical locations at the same time, so as to ensure that the preprocessed data can retain the characteristics of the original data and meet the input requirements of the prediction model.

[0023] Furthermore, in the calculation of the spatio-temporal interaction factor θ t,s of the multi-dimensional spatio-temporal sequence prediction module, the determination method of each influencing factor is further refined. For the interaction influence index δ 1 (t, s) between weather and power, it is determined by calculating the mutual information between the light intensity L s and the historical power P t . The mutual information I(L s , P t ) can measure the degree of dependence between the two. The formula is:

[0024]

[0025] Among them, p(l, p) is the joint probability distribution of the light intensity L and the power P, and p(l) and p(p) are their marginal probability distributions respectively. For the interaction influence index δ 2 (t, s) between wind speed and grid load, it is calculated by establishing a correlation function based on a physical model. Considering the influence of wind speed on wind power generation and the relationship between wind power generation and grid load, this correlation function can be expressed as δ 2 (t, s) = f(V s , G s ), where f(·) is a non-linear function established according to the wind power generation principle and grid operation characteristics. By comprehensively considering and quantifying multiple such influencing factors, the spatio-temporal interaction factor θ t,s can more accurately reflect the complex interaction relationship between the time and space dimensions, thereby improving the accuracy of the prediction model.

[0026] Furthermore, in the adaptive weight adjustment algorithm of the multi-dimensional spatio-temporal sequence prediction module, the determination methods of the learning rates β and γ are further optimized, and a dynamic learning rate adjustment strategy is adopted. According to the prediction error e l 's changing trend, the learning rate is dynamically adjusted. When the prediction error e l is large and shows an upward trend, the learning rate is appropriately increased to accelerate the speed of weight adjustment, so that the model can adapt to the data changes faster. When the prediction error e l is small and tends to be stable, the learning rate is decreased to avoid excessive weight adjustment, resulting in model instability. Specifically, the adjustment formula for the learning rate is:

[0027]

[0028]

[0029] where η is the learning rate adjustment coefficient, ∈ is the error change threshold, both determined through experiments and experience. Through this dynamic learning rate adjustment strategy, the convergence speed and stability of the adaptive weight adjustment algorithm can be improved, enabling the prediction model to reach the optimal state faster.

[0030] Furthermore, in the comprehensive optimization objective function F of the energy storage scheduling strategy formulation module, the calculation methods of each part and the determination basis of the weight coefficients are further elaborated. The calculation of the discharge revenue R takes into account the electricity price and discharge power in different time periods, and the formula is where t 1 and t 2 are the discharge time periods, P d (t) is the discharge power at time t. The calculation of the charging cost C takes into account the charging electricity price and charging power, and the formula is: where t 3 and t 4 are the charging time periods, P c (t) is the charging power at time t. The optimal state of charge SOC optimal of the battery is determined according to the long-term operation plan of the energy storage power station and the requirements of the power grid. The weight coefficients λ 1、λ2 and λ 3 are determined by the analytic hierarchy process. First, a hierarchical structure model is constructed, and the operation objectives of the energy storage power station are divided into multiple levels such as economic benefits, battery health management, and equipment reliability. Then, through expert evaluation and comparison of the relative importance between different objectives, a judgment matrix is constructed. Finally, the eigenvector of the judgment matrix is solved to obtain the values of each weight coefficient. In this way, the comprehensive optimization objective function can more comprehensively reflect the actual operation requirements of the energy storage power station and formulate a more reasonable scheduling strategy.

[0031] Furthermore, the execution and monitoring module adopts a distributed monitoring architecture. Multiple local monitoring nodes are arranged in different areas and key equipment of the energy storage power station. Each node is responsible for collecting and processing the operation data in its area. The local monitoring nodes transmit the processed data to the central monitoring server through a wireless communication network. The central monitoring server aggregates and analyzes the data from each local monitoring node to grasp the overall operation status of the energy storage power station in real time. At the same time, a local decision-making mechanism is set on the local monitoring nodes. When abnormal conditions are detected in local equipment, the local nodes can immediately take corresponding emergency measures to prevent the expansion of the fault. The central monitoring server evaluates the operation status of the entire energy storage power station based on the aggregated data. When it is found that the overall operation parameters deviate from the normal range, the scheduling strategy is adjusted in a timely manner, and the adjusted strategy is sent to each local monitoring node for execution. In addition, the execution and monitoring module also has a data backup and recovery function, regularly backing up the collected data and operation records to an external storage device to prevent data loss.

[0032] Furthermore, the energy storage power station management system based on EMS also includes an intelligent early warning module. This module is connected to the data acquisition module, the multi-dimensional spatio-temporal sequence prediction module, and the execution and monitoring module, and receives the data of each module in real time. The intelligent early warning module uses the anomaly detection algorithm in machine learning to monitor and analyze the operation data of the energy storage power station in real time. This algorithm identifies the anomaly points in the data by constructing a random forest. When it is detected that the value of a parameter significantly deviates from the normal range, it is determined as an abnormal situation. The intelligent early warning module also combines the prediction results of the multi-dimensional spatio-temporal sequence prediction module to give an early warning of potential risks that may occur. The warning signal is sent to the operation and maintenance personnel in various ways. At the same time, the intelligent early warning module also provides a detailed anomaly analysis report, including the time, location, and possible cause information of the anomaly, to help the operation and maintenance personnel quickly locate and solve problems, improving the operation safety and reliability of the energy storage power station.

[0033] Compared with the prior art, the energy storage power station management system based on EMS has the following beneficial effects:

[0034] First, this system collects comprehensive data in the time dimension and space dimension through the data acquisition module, and uses the multi-dimensional spatio-temporal sequence prediction module for accurate prediction, and can accurately predict the charge and discharge demand of the energy storage power station. The energy storage scheduling strategy formulation module formulates the optimal scheduling strategy according to the prediction results, realizing the efficient operation of the energy storage power station. This can not only maximize the utilization of the capacity of the energy storage power station, improve the charge and discharge efficiency, but also formulate discharge and charge plans according to the electricity prices in different time periods, thereby reducing the operation cost and increasing the economic benefits.

[0035] Second, the system identifies outliers through the construction of a random forest machine learning algorithm to promptly warn of potential risks. In addition, the execution and monitoring module adopts a distributed monitoring architecture, which can quickly take emergency measures when detecting abnormal situations to prevent the expansion of faults. This comprehensive monitoring and warning mechanism can significantly improve the operating safety of the energy storage power station. At the same time, the system can also adjust and optimize the prediction model and scheduling strategy in real time to ensure that the energy storage power station always maintains the best operating state, further enhancing its operating reliability.

[0036] Other advantages, objectives, and features of the present invention will be described to some extent in the subsequent specification, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0038] Figure 1 It is a flowchart of the function implementation of an energy storage power station management system based on EMS;

[0039] Figure 2 It is an overall architecture diagram of an energy storage power station management system based on EMS. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] To further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention objective, the following will, in conjunction with the accompanying drawings and preferred embodiments, detail the specific implementation manners, structures, features, and their effects of the present invention as follows.

[0041] Embodiment 1

[0042] This embodiment describes that an energy storage power station is built in the fringe area of a certain city, mainly supplying power to the surrounding industrial parks and residential areas. The peak electricity consumption period in this area is concentrated from 9:00 to 12:00 and from 18:00 to 22:00 on weekdays, and the low valley period is from 0:00 to 6:00 in the early morning.

[0043] Collect historical power data P at a high sampling frequency of 15 minutes during the peak electricity consumption period t , and adopt a low sampling frequency of 1 hour during the low valley period. At the same time, collect electricity price fluctuation data E t , and construct a background knowledge base for electricity price prediction, and collect weather difference data W s(including humidity H s , air pressure P s ), obtain the surrounding power grid load distribution data G s by interacting with multiple power grid dispatching centers, and conduct quality assessment on the data.

[0044] Preprocess the collected data. For power data, extract mean, variance, and peak features. For weather data, extract correlation features and then perform clustering processing. Use adaptive normalization for power data and linear normalization for some meteorological data. When constructing the prediction model, calculate the spatio-temporal interaction factor θ t,s , such as calculating the mutual information between the light intensity L s and the historical power P t to determine the weather and power interaction influence index δ (t, s), and the prediction formula is 1 (t, s), and the prediction formula is where The energy storage dispatching strategy formulation module formulates strategies according to the prediction result Y t+1,s , combined with the state of charge SOC of the battery t , charge and discharge power limits P max , P min and the healthy operation state H of the equipment, and uses the comprehensive optimization objective function F = λ 1 ·(R - C)+λ 2 ·(SOC t+1 - SOC optimal ) 2 +λ 3 ·H to formulate strategies, where the discharge income the charging cost and the weight coefficients λ 1 , λ 2 , λ 3 are determined by the analytic hierarchy process.

[0045] Adopt a distributed monitoring architecture. Local monitoring nodes collect and process data and upload it. The central monitoring server aggregates and analyzes it. When local equipment anomalies are detected, local nodes perform emergency processing. When overall operation parameter anomalies occur, the central server adjusts the dispatching strategy. The intelligent early warning module monitors and analyzes data in real time, detects anomalies through the random forest algorithm, combines the prediction results for early warning, and notifies the operation and maintenance personnel.

[0046] Embodiment 2

[0047] This embodiment describes that in remote mountainous areas with complex terrain and changeable climate, severe weather occurs frequently. Strong winds may damage energy storage equipment, heavy rains are likely to cause floods threatening the safety of power stations, and cloudy days will reduce the solar power generation efficiency. In such an environment, the energy storage power station not only needs to maintain its own stable operation, but also needs to ensure the power supply in surrounding areas and reduce the negative impact of weather on the power system.

[0048] Strengthen the monitoring of meteorological parameters in bad weather, increase the monitoring frequency of light intensity and wind speed to once every 5 minutes, and adjust the acquisition frequency of historical power data in real time according to weather and electricity load fluctuations. For example, when the weather suddenly changes, it is increased to once a minute. By interacting with surrounding meteorological stations, power grid monitoring points and users, collect comprehensive power grid load distribution data and promptly eliminate abnormal data.

[0049] Optimize the processing method according to the data characteristics in bad weather. When preprocessing the data, in addition to extracting conventional features from the power data, calculate the dynamic feature of the change rate, and analyze the non-linear correlation of the meteorological data. According to these features, subdivide the data categories and adopt targeted normalization methods for different categories. For example, use sliding window adaptive normalization for power data with large fluctuations.

[0050] When calculating the spatio-temporal interaction factor, expand the influencing factors. In addition to the mutual information between light intensity and historical power, construct a more complex correlation function between wind speed and power grid load, incorporate the influence of terrain on wind speed, and at the same time consider the synergistic effect of various bad weathers to improve the prediction accuracy.

[0051] According to the prediction results, formulate strategies in combination with battery status, charge and discharge power limits, and equipment health conditions. For example, when it is predicted that there will be a long period of cloudy days, appropriately reduce discharge and increase charging to maintain the stability of the battery power. By dynamically adjusting the weight coefficients of the comprehensive optimization objective function, adapt to the operation requirements under different bad weathers. For example, during heavy rain, increase the weight coefficient corresponding to the equipment health status to ensure equipment safety.

[0052] Adopt a distributed architecture, arrange local monitoring nodes in each area and key equipment of the power station to collect equipment operation data in real time, such as battery voltage and the working status of the inverter, and transmit it to the central monitoring server through wireless communication network. The local monitoring node has a local decision-making mechanism and can immediately take measures when abnormal equipment is found. For example, when the battery temperature is too high, start heat dissipation and power reduction. The central monitoring server evaluates the overall status of the power station according to the aggregated data. When it deviates from the normal range, adjust the dispatching strategy in time and issue an execution order. In addition, this module also regularly backs up data to external storage devices to ensure data security.

[0053] Receive the data of each module in real time, use the random forest algorithm to monitor the operation data. Once the parameters deviate from the normal range, such as abnormal battery charging voltage and excessive wind turbine speed, immediately alarm. Combine the prediction results to give early warnings of potential risks in advance, such as reminding to protect the wind turbine before strong winds come. Send warning signals to the operation and maintenance personnel in various ways and provide a detailed abnormal analysis report to help solve problems quickly.

[0054] The above are only the preferred embodiments of the present invention, and do not impose any form of limitation on the present invention. Although the present invention has been disclosed above with the preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments by using the above-disclosed technical content within the scope of the technical solution of the present invention. However, as long as it does not depart from the content of the technical solution of the present invention, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. An energy storage power station management system based on EMS, characterized in that: The system includes the following components: data acquisition module, multi-dimensional spatiotemporal series prediction module, energy storage scheduling strategy formulation module and execution and monitoring module; The data acquisition module collects historical power data and electricity price fluctuation data in the time dimension, as well as weather difference data in different geographical locations in the spatial dimension, and surrounding power grid load distribution data. The data is collected in real time through sensors, meteorological stations, and power grid monitoring equipment distributed in different locations; The multi-dimensional spatiotemporal sequence prediction module pre-processes the collected data and unifies different types of data into the same dimension. Let the pre-processed data vector be X t,s =[P t ,E t ,W s ,G s ], construct a multidimensional spatiotemporal sequence prediction model, and introduce a new spatiotemporal interaction factor θ t,s , which reflects the degree of mutual influence between time t and geographical location s. The prediction model is based on the following formula: in, is the predicted value of the charging and discharging demand of the energy storage power station at time t+1 and geographical location s. n and m are the historical data backtracking steps in time and space dimensions, respectively. i,j is the weight coefficient, indicating that historical data X t-i,s-j The degree of influence on the predicted value, f(·) is a nonlinear transformation function, using Where k is adaptively adjusted according to the battery aging degree; Spatiotemporal interaction factor θ t,s The calculation method is: Among them, q is the number of factors that affect the spatiotemporal interaction, α k is the weight of each factor, δ k (t, s) is the influence index of the kth factor at time t and geographical location s. When k = 1, it means the interaction between weather and power. δ1(t, s) can be calculated by the light intensity L s and historical power P t The correlation of is calculated; The energy storage scheduling strategy formulation module: based on the charge and discharge demand prediction results output by the multi-dimensional spatiotemporal sequence prediction module Combined with the current battery power state SOC of the energy storage power station t , Battery charge and discharge power limit P max and P min , the operating health status H of the equipment, formulate the energy storage scheduling strategy, and use a comprehensive optimization objective function F to determine the optimal scheduling strategy: Among them, R is the discharge benefit, C is the charging cost, SOC optimal is the optimal state of charge of the battery, λ1, λ2 and λ3 are weight coefficients, which are determined by evaluating the importance of different operating objectives of the energy storage power station. The optimization objective function is solved by the gradient descent method to obtain the optimal charging and discharging time and power scheduling parameters; The execution and monitoring module is responsible for executing the scheduling strategy generated by the energy storage scheduling strategy formulation module, controlling the operation of the charging and discharging equipment of the energy storage power station, and at the same time, monitoring the operating status of the energy storage power station in real time, and feeding back these data to the data acquisition module and the multi-dimensional space-time series prediction module, so as to adjust and optimize the prediction model and scheduling strategy in real time.

2. The EMS-based energy storage power station management system according to claim 1, characterized in that: The data acquisition module, for historical power data P t The multi-resolution sampling strategy is used to collect data during peak power consumption periods at a higher sampling frequency to capture rapid changes in power. During low power consumption periods, a lower sampling frequency is used to collect data on electricity price fluctuations. t , and also collects historical electricity price adjustment policies and trend information of the electricity market to build a background knowledge base for electricity price forecasting. s In terms of collection, the humidity H s , air pressure P s Meteorological parameter collection, surrounding power grid load distribution data G s By exchanging data with multiple power grid dispatch centers at different levels, more comprehensive load information can be obtained. At the same time, a preliminary quality assessment of the collected data is performed to eliminate data with obvious errors or abnormalities.

3. The EMS-based energy storage power station management system according to claim 1, characterized in that: The data preprocessing part of the multidimensional spatiotemporal series prediction module adopts a preprocessing method based on data feature clustering. First, the features of different types of collected data are extracted. For the power data P t , extract its mean, variance, and peak statistical features, for weather data W s ,The correlation characteristics between different meteorological parameters are extracted, and then the data are clustered and divided into different categories based on these characteristics. For each category, different normalization methods are used for processing. For data categories with large fluctuations, such as power data, an adaptive normalization method is used to dynamically adjust the normalization scale according to the real-time fluctuation range of the data. For relatively stable data categories, traditional linear normalization methods are used for some meteorological data.

4. The EMS-based energy storage power station management system according to claim 1, characterized in that: The spatiotemporal interaction factor θ of the multidimensional spatiotemporal sequence prediction module t,s In the calculation, the determination method of each influencing factor is further refined. For the interactive influence index δ1(t,s) of weather and power, the light intensity L is calculated. s and historical power P t The mutual information I(L s ,P t ) can measure the degree of dependence between the two, the formula is: Among them, p(l,p) is the joint probability distribution of light intensity L and power P, p(l) and p(p) are their marginal probability distributions respectively. For the interactive impact index of wind speed and grid load δ2(t,s), a correlation function based on a physical model is established to calculate. Considering the impact of wind speed on wind power generation and the relationship between wind power generation and grid load, the correlation function can be expressed as δ2(t,s)=f(V s ,G s ), where f(·) is a nonlinear function established according to the wind power generation principle and the grid operation characteristics. By comprehensively considering and quantifying multiple such influencing factors, the spatiotemporal interaction factor θ t,s It can more accurately reflect the complex interactive relationship between time and space dimensions, thereby improving the accuracy of the prediction model.

5. The EMS-based energy storage power station management system according to claim 1, characterized in that: In the adaptive weight adjustment algorithm of the multidimensional spatiotemporal sequence prediction module, the determination method of the learning rates β and γ is further optimized, and a dynamic learning rate adjustment strategy is adopted to adjust the learning rate according to the prediction error e l The learning rate is adjusted dynamically according to the changing trend of the prediction error e l When the prediction error e is large and on an upward trend, the learning rate should be appropriately increased to speed up the weight adjustment so that the model can adapt to data changes more quickly. l When it is small and tends to be stable, reduce the learning rate to avoid excessive weight adjustment, which may lead to model instability. Specifically, the learning rate adjustment formula is: Among them, η is the learning rate adjustment coefficient, and ∈ is the error change threshold.

6. The EMS-based energy storage power station management system according to claim 1, characterized in that: In the comprehensive optimization objective function F of the energy storage scheduling strategy formulation module, the calculation method of each part and the basis for determining the weight coefficient are further described in detail. The calculation of the discharge benefit R takes into account the electricity price and discharge power in different time periods. The formula is: Where t1 and t2 are the discharge time periods, P d (t) is the discharge power at time t. The calculation of charging cost C takes into account the charging price and charging power, and the formula is: Where t3 and t4 are charging time periods, P c (t) is the charging power at time t, and the optimal state of charge SOC of the battery optimal The weight coefficient λ is determined according to the long-term operation plan of the energy storage power station and the needs of the power grid. 1、λ2 and λ3 are determined by analytic hierarchy process.

7. The EMS-based energy storage power station management system according to claim 1, characterized in that: The execution and monitoring module adopts a distributed monitoring architecture, and multiple local monitoring nodes are arranged in different areas and key equipment of the energy storage power station. Each node is responsible for collecting and processing the operating data in the area. The local monitoring node transmits the processed data to the central monitoring server through the wireless communication network. The central monitoring server summarizes and analyzes the data from each local monitoring node, and grasps the overall operating status of the energy storage power station in real time. At the same time, a local decision-making mechanism is set on the local monitoring node. When an abnormal situation is detected in a local device, the local node can immediately take corresponding emergency measures. The central monitoring server evaluates the operating status of the entire energy storage power station based on the summarized data. When it is found that the overall operating parameters deviate from the normal range, the scheduling strategy is adjusted in time, and the adjusted strategy is sent to each local monitoring node for execution. In addition, the execution and monitoring module also has data backup and recovery functions, and regularly backs up the collected data and operation records to external storage devices.

8. The EMS-based energy storage power station management system according to claim 1, characterized in that: The EMS-based energy storage power station management system also includes an intelligent early warning module, which is connected to the data acquisition module, the multidimensional space-time sequence prediction module and the execution and monitoring module, and receives data from each module in real time. The intelligent early warning module uses the anomaly detection algorithm in machine learning to monitor and analyze the operation data of the energy storage power station in real time. The algorithm identifies anomalies in the data by constructing a random forest. When it is detected that the value of the parameter deviates significantly from the normal range, it is determined to be an abnormal situation. The intelligent early warning module also combines the prediction results of the multidimensional space-time sequence prediction module to provide early warning of potential risks that may arise. The warning signal is sent to the operation and maintenance personnel in a variety of ways. At the same time, the intelligent early warning module also provides a detailed abnormal analysis report, including the time, location, possible causes and other information of the abnormality, to help the operation and maintenance personnel quickly locate and solve the problem.

Images