Energy storage power station monitoring device

Abstract

The invention relates to the technical field of energy storage operation and maintenance management, in particular to an energy storage power station monitoring device. Comprising a multi-parameter monitoring module, a data processing unit, a communication module and a display and alarm unit, the multi-parameter monitoring module collects voltage, current and environmental parameters of the battery pack by integrating a high-sensitivity sensor array; an intelligent algorithm is built in the data processing unit, abnormal conditions in operation are rapidly recognized by analyzing monitoring data in real time, and an optimized operation scheme is generated in combination with historical data; the communication module supports wireless and wired transmission; and the display and alarm unit is used for locally displaying key parameters and reminding abnormal conditions. According to the energy storage power station monitoring system, high-precision real-time monitoring and operation optimization of the energy storage power station are realized by integrating the synergistic effect of multiple modules, and the problems of low monitoring precision, insufficient safety performance and lack of intelligent analysis in a traditional energy storage power station monitoring system are effectively solved, so that the operation efficiency, the safety and the management level of the energy storage power station are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage operation and maintenance management, and in particular to a monitoring device for an energy storage power station. Background Art

[0002] As an important part of the modern power system, energy storage power stations play a vital role in electric energy storage and regulation, grid load balancing, and efficient use of renewable energy. However, with the expansion of energy storage scale and the increase in technical complexity, traditional monitoring systems have gradually exposed problems such as insufficient monitoring accuracy, weak response capabilities, low intelligence level, and poor environmental adaptability. For example, existing systems can usually only monitor basic parameters such as voltage and current of battery cells, lack a comprehensive assessment of the operating status and performance of the overall energy storage power station, and are difficult to detect and warn of abnormal situations in a timely manner, posing a safety hazard. At the same time, most monitoring equipment only has data acquisition functions, fails to achieve intelligent operation optimization, and shows insufficient earthquake resistance, dustproofness, and waterproofness in complex environments, which increases the difficulty of maintenance.

[0003] In the prior art, for example, the invention patent with the authorization announcement number CN116455085A provides a battery energy storage power station intelligent monitoring system, which is used to realize the intelligent monitoring of the battery energy storage power station and improve the accuracy of monitoring operation and maintenance. The system includes: a battery cell status monitoring module, a module temperature monitoring module and a power station fire alarm module; the battery cell status monitoring module is used to: obtain the battery cell body parameters and monitoring parameters of the battery energy storage power station to monitor the battery cell status, and output multiple battery cell status detection results corresponding to each battery cell status analysis model; the module temperature monitoring module is used to: obtain the array module temperature of the battery energy storage power station to perform module temperature early warning analysis, and generate module temperature early warning analysis results; the power station fire alarm module is used to: obtain the energy storage power station operating data of the battery energy storage power station to perform power station fire alarm analysis, and obtain the power station fire alarm analysis results.

[0004] For example, the invention patent with the authorization announcement number CN112085930A provides a generator set intelligent monitoring and early warning system and method, including a data acquisition module, a server and an intelligent early warning module; the data acquisition module is used to collect the physical quantity information of vibration, swing, temperature, voltage, current and key phase / speed of the generator set, and transmit it to the server in real time; the server is used to automatically analyze and store the physical quantity information of the real-time state of the generator set and the physical quantity information of the historical state; the intelligent early warning module is used to classify and warn according to the results of the automatic analysis. Through online monitoring of the physical quantity information of the generator set, abnormal and fault conditions of the unit during operation can be discovered in time, and an alarm signal can be issued according to the pre-set logic and fixed value to carry out safety warning, so as to achieve the purpose of protecting the safe operation of the unit; and it can intelligently analyze the health status of the generator set, with high versatility and wide application range. However, the above scheme mainly focuses on the voltage and current monitoring of the battery unit, lacks the comprehensive monitoring and management function of the overall energy storage power station, and is difficult to meet the intelligent needs of modern energy storage power stations. In view of this, we propose a monitoring device for energy storage power stations. Summary of the invention

[0005] The purpose of the present invention is to provide a monitoring device for an energy storage power station to solve the problems raised in the above background technology.

[0006] In order to solve the above technical problems, the present invention aims to provide a monitoring device for an energy storage power station, comprising a multi-parameter monitoring module, a data processing unit, a communication module and a display and alarm unit which are sequentially connected in communication; wherein:

[0007] The multi-parameter monitoring module collects the voltage, current and environmental parameters of the battery pack by integrating a high-sensitivity sensor array to ensure the comprehensiveness and accuracy of the data;

[0008] The data processing unit has a built-in intelligent algorithm that can quickly identify abnormal conditions (such as overheating, overvoltage or failure) during operation by analyzing monitoring data in real time, and generate optimized operation plans based on historical data to achieve predictive maintenance and fault warning;

[0009] The communication module supports wireless and wired transmission, and is connected to the cloud management platform. Operation and maintenance personnel can remotely view the device status and receive abnormal alarms, further improving management efficiency;

[0010] The display and alarm unit displays key parameters through an LCD screen and is equipped with an audible and visual alarm function to give an immediate warning when an abnormal situation occurs.

[0011] As a further improvement of the technical solution, in the multi-parameter monitoring module, the high-sensitivity sensor array includes at least a voltage sensor, a current sensor, a temperature sensor, an ambient humidity sensor, an air pressure sensor, a combustible gas sensor and a composite detector;

[0012] The voltage sensor is used to collect the voltage parameters of the battery pack; the current sensor is used to accurately measure the current size of the battery pack; the temperature sensor is responsible for monitoring the temperature changes of the battery pack; the environmental humidity sensor is used to collect humidity information in the energy storage power station environment; the air pressure sensor is used to obtain the environmental air pressure parameters; the combustible gas sensor is used to achieve early fire warning and ensure the safety of the power station environment; the composite detector is used to achieve comprehensive monitoring of multiple disasters and link the fire protection system.

[0013] As a further improvement of the present technical solution, the environmental parameters that need to be collected by the multi-parameter monitoring module include temperature, humidity and air pressure.

[0014] As a further improvement of the technical solution, the operation process of the data processing unit includes the following steps:

[0015] S2.1. Data collection: Collect the operating data of various equipment in the energy storage power station;

[0016] S2.2, Data preprocessing: Cleaning, format conversion and standardization of the collected raw data;

[0017] S2.3, Data storage: Store the preprocessed data in an appropriate location for subsequent use;

[0018] S2.4, Data processing and analysis: Further processing and analysis of stored data, mainly including data processing, data analysis and data visualization;

[0019] S2.5, Data application: Apply the processed and analyzed data to the actual operation of the energy storage power station;

[0020] S2.6, Security and Maintenance: An important link to ensure the stable operation of the data processing unit and protect data security, including data security management, system maintenance and troubleshooting.

[0021] As a further improvement of the technical solution, in step S2.2, preprocessing the data includes:

[0022] Data cleaning is used to remove noise, outliers and duplicate data in the data to ensure the accuracy and consistency of the data. Data cleaning includes two types: outlier processing and noise data smoothing. Outlier processing is based on the 3σ principle, and the specific algorithm is:

[0023] Assume that the monitoring data sequence is x1,x2,...,xn , first calculate the mean of the series and standard deviation σ;

[0024] Mean calculation formula:

[0025]

[0026] Standard deviation calculation formula:

[0027]

[0028] Then determine whether the data point is an outlier. If x i satisfy The data point is considered an outlier and can be processed using the following methods:

[0029] a. Delete outliers directly: i Delete from the data series to obtain a new data series for subsequent analysis;

[0030] b. Correct outliers: You can use the mean Or other reasonable values ​​to replace x i , making the data series smoother and more reasonable;

[0031] The moving average method is used for noise data smoothing. The specific algorithm is as follows:

[0032] For data sequences with noise, the moving average method is used for smoothing; let the window size be m (m is an odd number, and m <n);

[0033] For data points The smoothed result y i The calculation formula is:

[0034]

[0035] That is, in x i Take both before and after The data points are averaged as x i The smoothing value of

[0036] Data format conversion is used to convert data from different sources and formats into a unified format for subsequent processing and analysis; data format conversion includes analog to digital conversion (A / D conversion), conversion between different data types, and data encoding format conversion;

[0037] Data normalization is used to scale or transform data according to specific specifications for easy comparison and analysis; data normalization includes minimum-maximum normalization, Z-score normalization, and decimal scaling normalization.

[0038] As a further improvement of the present technical solution, the intelligent algorithms built into the data processing unit include at least a data anomaly detection algorithm, a fault prediction algorithm and an operation optimization algorithm; the data anomaly detection algorithm includes a threshold-based algorithm, a statistical analysis algorithm and a machine learning algorithm; the fault prediction algorithm includes a time series analysis algorithm and a neural network algorithm; the operation optimization algorithm includes a dynamic programming algorithm and a model predictive control algorithm.

[0039] As a further improvement of the technical solution, the specific algorithm process of the data anomaly detection algorithm includes:

[0040] First, the threshold-based algorithm: set the normal threshold range of parameters such as voltage, current, temperature, humidity, air pressure, combustible gas content or smoke concentration. When the monitoring data exceeds this range, it is judged as abnormal; the algorithm formula is:

[0041] Let the monitoring parameter be x and the lower limit of the normal threshold be x min , the upper limit of the normal threshold is x max ;

[0042] If x <x min or x>x max , then the data x is judged to be abnormal;

[0043] Second, statistical analysis algorithm: using the mean-standard deviation method, by calculating the mean and standard deviation of the monitoring data over a period of time, determine the distribution range of normal data; if the newly collected data deviates from this range to a certain extent, it is considered abnormal; the algorithm process is:

[0044] Assume that the monitoring data sequence is {x1,x2,...,x n};

[0045] Calculate the mean

[0046] Calculate standard deviation

[0047] Determine the interval of normal data as Where k is a coefficient set based on experience or actual conditions, usually 2 or 3;

[0048] if or The data is judged to be abnormal;

[0049] Third, machine learning algorithm: clustering algorithm is used to cluster the monitoring data under normal operation to form normal data clusters; when the new data point is far away from these normal clusters, it is judged as abnormal; K-Means algorithm is used, and its algorithm process is as follows:

[0050] Assume that the monitoring data is an n-dimensional vector set X = {x1, x2, ..., x m}, where x i ={x i1 ,x i2 ,...,x in};

[0051] Randomly select K initial cluster centers C = {c1, c2, ..., c K}, where c j ={c j1 ,c j2 ,...,c jn}, j = 1, 2, ..., K;

[0052] For each data point x i , calculate it to each cluster center c j The distance d ij , usually using the Euclidean distance formula

[0053] The data point x i Assign to the cluster to which the nearest cluster center belongs;

[0054] Recalculate the cluster center of each cluster, that is, for the jth cluster, the new cluster center c j The value of each dimension of is the average value of the corresponding dimension of all data points in the cluster, that is, Where N j is the number of data points in the jth cluster;

[0055] Repeat the above steps of assigning data points and updating cluster centers until the cluster centers no longer change significantly (converge);

[0056] For a new data point x new , calculate its distance to each cluster center, if its distance to the nearest cluster center is greater than a certain set threshold, then determine x new is abnormal.

[0057] As a further improvement of the technical solution, the specific algorithm process of the fault prediction algorithm includes:

[0058] First, time series analysis algorithm: Use the ARIMA model to model and analyze historical monitoring data (such as the time series of the voltage and current of the battery pack) to predict the trend of future data; if the predicted value deviates greatly from the actual monitoring value, or the predicted trend shows that abnormal changes are about to occur (such as the continuous downward trend of voltage may indicate that the battery capacity is declining or is about to fail), an early warning fault is issued; the specific algorithm process includes:

[0059] Model building process:

[0060] Assume that the monitoring data sequence is {y t}, t = 1, 2, ..., T;

[0061] First, perform a stationary test on the data; if the data is non-stationary, differential processing is required until the data is stable; let the number of differentials be d;

[0062] Determine the autoregressive order p and moving average order q of the ARIMA model;

[0063] The formula for establishing the ARIMA (p, d, q) model is:

[0064]

[0065] in is an autoregressive operator, θ(B)=1+θ1B+θ2B 2 +...+θ q B q is the moving average operator, and B is the lag operator (i.e., B k y t =y t -k),∈ t is a white noise sequence;

[0066] Prediction process:

[0067] Estimate the parameters of the ARIMA model and θ1,θ2,...,θ q , we can use methods such as least squares or maximum likelihood estimation;

[0068] The predicted value for the future time T+h, (h=1,2,...) It can be obtained by recursive calculation;

[0069] If the predicted value Compared with the actual monitoring value y T+h If the deviation is large, or the forecast trend shows that an abnormal change is about to occur, an early warning of the failure will be given;

[0070] Second, neural network algorithm: using recurrent neural network RNN ​​and its variant LSTM, by learning a large amount of historical data, a complex nonlinear relationship model between the battery pack operating status and various monitoring parameters is established; the neural network can capture the long-term dependencies and hidden patterns in the data, thereby predicting the probability of failure; the specific algorithm process includes:

[0071] RNN basic structure and algorithm process:

[0072] The basic structure of RNN includes input layer, hidden layer and output layer; let the input sequence be {xt}, t=1,2,...,T, the hidden layer state is {h t}, the output is {y t};

[0073] The update formula of the hidden layer state is:

[0074] h t =f(Ux t +Wh t-1 +b)

[0075] Where U is the weight matrix from the input layer to the hidden layer, W is the weight matrix from the hidden layer to the hidden layer (self-circular connection), b is the bias vector of the hidden layer, and f(·) is the activation function, such as tanh or ReLU;

[0076] The output formula of the output layer is:

[0077] y t =g(Vh t +c)

[0078] Where V is the weight matrix from the hidden layer to the output layer, c is the bias vector of the output layer, and g(·) is the activation function, such as softmax.

[0079] During training, the back-propagation algorithm is used to update the weight matrices U, W, and V and the bias vectors b and c; the goal is to minimize the loss function, such as the mean squared error loss function:

[0080]

[0081] in is the true output value;

[0082] Use LSTM to improve RNN, LSTM structure and algorithm process:

[0083] LSTM improves the unit structure of the hidden layer based on RNN; the LSTM unit contains input gate, forget gate and output gate;

[0084] The formula of the forget gate is:

[0085] f t =σ(W f x t +U f h t-1 +b f )

[0086] Where W f , U f is the corresponding weight matrix, b fis the bias vector, σ(·) is the sigmoid function;

[0087] The formula for the input gate is:

[0088] i t =σ(W i x t +U i h t-1 +b i )

[0089] And there is a new candidate state

[0090] The formula for the output gate is:

[0091] o t =σ(W o x t +U o h t-1 +b o )

[0092] The update formula of the hidden layer state is:

[0093]

[0094] The output formula of the output layer is similar to that of RNN;

[0095] After training the LSTM model with a large amount of historical data, the new monitoring data is input into the model to obtain a predicted output of the probability of fault occurrence; if the predicted probability exceeds the set threshold, a fault warning is issued.

[0096] As a further improvement of the technical solution, the specific algorithm process of running the optimization algorithm includes:

[0097] First, dynamic programming algorithm: Taking into account the operating constraints of the energy storage power station (such as battery pack charging and discharging power limits, energy storage capacity limits, etc.) and goals (such as minimizing operating costs, maximizing energy utilization efficiency, etc.), the dynamic programming method is used to optimize the battery pack charging and discharging strategy; the algorithm process includes:

[0098] Assume that the operation cycle of the energy storage power station is divided into T time periods, each of which is represented by t=1, 2, ..., T;

[0099] Define the state variable S t Indicates the status of the energy storage power station at the beginning of time period t, including information such as the remaining power of the battery pack;

[0100] Define the decision variable x t represents the decision taken at time period t, for example, the battery pack is charged (x t =1), discharge (xt = -1) or standby (x t =0);

[0101] Define the cost function C(S t ,x t ) indicates that in state S t Take decision x t the costs incurred;

[0102] Define the transfer function S t+1 =T(S t ,x t ) indicates that according to the current state S t and decision x t , get the state at the beginning of the next time period t+1;

[0103] The dynamic programming algorithm solves the optimal decision sequence by following these steps:

[0104] Initialization: For t = T, calculate the terminal cost That is, find the decision that minimizes the cost in the last time period;

[0105] Recursive calculation: For t = T-1, T-2, ..., 1, calculate That is, considering the current cost and the optimal cost of the next time period, find the decision that minimizes the total cost in the current time period;

[0106] Optimal decision sequence: By backtracking the above calculation process, we can get the optimal decision sequence from t = 1 to t = T This sequence is the charging and discharging strategy that minimizes the cost of the energy storage power station throughout the entire operating cycle.

[0107] Second, the model predictive control algorithm (MPC): Based on the dynamic model of the system, in each control cycle, the optimal control input is calculated according to the current monitoring data and the predicted future system state; the algorithm process includes:

[0108] Assume that the dynamic model of the energy storage power station can be expressed by the following discrete time state space equation:

[0109] x k+1 =f(x k ,u k )+w k

[0110] y k =g(x k )+v k

[0111] where x k is the state vector of the system at time k, uk is the control input vector, y k is the measured output vector, f(·) and g(·) are the dynamic functions of the system, and w k and v k are process noise and measurement noise respectively;

[0112] In each control cycle k, the MPC algorithm performs the following steps:

[0113] Prediction model: Based on the state x at the current time k k and the measured output y k , and the predicted future system state, using the system's dynamic model to predict the system state for N time steps in the future and measurement output in represents the state at time k+i predicted at time k, represents the measured output at time k+i predicted at time k;

[0114] Objective function setting: Define the objective function Where L(·) is the loss function that measures the system performance, and M(·) is the function that measures the control input cost;

[0115] Optimization solution: By solving the minimum value of the objective function J, the optimal control input sequence for the next N-1 time steps is obtained. That is, to find the charging and discharging power instruction sequence that optimizes system performance and control cost in the future;

[0116] Implement control: Input the first optimal control Apply to the system and then repeat the above steps in the next control cycle.

[0117] As a further improvement of the technical solution, it also includes a structural design module, which adopts a modular earthquake-resistant, dust-proof and waterproof design to ensure the stable operation of the device in a complex environment.

[0118] Compared with the prior art, the present invention has the following beneficial effects:

[0119] 1. The energy storage power station monitoring device is based on an intelligent monitoring system with multiple modules working together. By integrating a multi-parameter monitoring module, a data processing unit, a communication module, a display and alarm unit, and a structural design module, it can achieve high-precision real-time monitoring and operation optimization of the energy storage power station, effectively solving the problems of low monitoring accuracy, insufficient safety performance, and lack of intelligent analysis in the traditional energy storage power station monitoring system, thereby improving the operation efficiency, safety, and management level of the energy storage power station;

[0120] 2. The energy storage power station monitoring device can improve the accuracy and efficiency of monitoring the operating status of energy storage battery packs; through real-time data analysis and alarm functions, it can significantly improve the safety of energy storage power stations; support intelligent operation optimization, reduce operation and maintenance costs, and extend the service life of battery packs; at the same time, the modular design facilitates installation and maintenance and adapts to the needs of different energy storage power stations; through synergy, it not only improves the operating efficiency and safety of energy storage power stations, but also realizes intelligent management and efficient operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0121] Figure 1 is an exemplary overall device architecture block diagram of the present invention;

[0122] Figure 2 It is an exemplary overall structural diagram of the present invention;

[0123] Figure 3 It is a detailed diagram of the internal structure of an exemplary multi-parameter monitoring module of the present invention;

[0124] Figure 4 is a schematic diagram of an exemplary operation flow of a data processing unit in the present invention;

[0125] Figure 5 It is a schematic diagram of the working principle of an exemplary display and alarm unit in the present invention;

[0126] In the figure: 100, multi-parameter monitoring module; 200, data processing unit; 300, communication module; 400, display and alarm unit; 500, structural design module. DETAILED DESCRIPTION

[0127] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0128] Example 1

[0129] like Figure 1-Figure 5 As shown, this embodiment provides an energy storage power station monitoring device, including a multi-parameter monitoring module 100, a data processing unit 200, a communication module 300 and a display and alarm unit 400 which are sequentially connected in communication; wherein:

[0130] The multi-parameter monitoring module 100 collects the voltage, current and environmental parameters of the battery pack by integrating a high-sensitivity sensor array to ensure the comprehensiveness and accuracy of the data;

[0131] The data processing unit 200 has a built-in intelligent algorithm that can quickly identify abnormal conditions (such as overheating, overvoltage or failure) during operation by analyzing monitoring data in real time, and generate optimized operation plans based on historical data to achieve predictive maintenance and fault warning;

[0132] The communication module 300 supports wireless and wired transmission, and is connected to a cloud management platform or a remote server. Operation and maintenance personnel can remotely view the device status and receive abnormal alarms, further improving management efficiency.

[0133] The display and alarm unit 400 displays key parameters on an LCD screen and is equipped with an audible and visual alarm function for local display of key parameters and reminder of abnormal conditions. When an abnormal condition occurs, an immediate warning is given, such as Figure 5 Shown is a schematic diagram of an exemplary working principle of the display and alarm unit 400.

[0134] In this embodiment, Figure 3 As shown, in the multi-parameter monitoring module 100, the high-sensitivity sensor array includes at least a voltage sensor, a current sensor, a temperature sensor, an ambient humidity sensor, an air pressure sensor, a combustible gas sensor, and a composite detector;

[0135] The voltage sensor is used to collect the voltage parameters of the battery pack. By monitoring the voltage value in real time, it can help determine the charging and discharging status of the battery and whether there are voltage anomalies, such as overvoltage or undervoltage, which is crucial for evaluating the performance and safety of the battery pack.

[0136] The current sensor is used to accurately measure the current of the battery pack. The current data obtained by the current sensor can be used to calculate the charge and discharge power of the battery pack. On the other hand, it can also be used to analyze the working efficiency of the battery and detect whether there is an abnormal current such as overcurrent, so as to timely discover potential circuit problems;

[0137] The temperature sensor is responsible for monitoring the temperature changes of the battery pack. The battery will generate heat during the charging and discharging process. Excessive temperature may affect the performance and life of the battery, and even cause safety problems. The temperature sensor can monitor the battery temperature in real time. When the temperature exceeds the set threshold, it will promptly trigger an alarm and corresponding processing mechanism to ensure the safe and stable operation of the battery pack.

[0138] The environmental humidity sensor is used to collect humidity information in the energy storage power station environment. Environmental humidity may affect the operation of the equipment. For example, excessive humidity may cause condensation on the surface of the equipment, which in turn affects electrical performance or causes corrosion. By monitoring humidity data, dehumidification and other measures can be taken in time to maintain the normal operation environment of the equipment.

[0139] The air pressure sensor is used to obtain ambient air pressure parameters. In some cases, air pressure changes may be associated with the operating status of the equipment. For example, in some sealed environments, abnormal air pressure may indicate problems with the sealing performance of the equipment or potential gas leakage risks, which helps to fully understand the operating environment of the energy storage power station.

[0140] Combustible gas sensors are used to achieve early fire warning and ensure the environmental safety of power stations: In energy storage power stations, battery packs may cause electrolyte leakage and other problems due to overheating, short circuit and other faults during operation. Organic solvents and other components in the electrolyte may form flammable gases after volatilization; combustible gas sensors can monitor the concentration of flammable gases in the environment in real time, and issue warning signals in time when the concentration reaches a certain threshold; for example, when the concentration of flammable gases such as hydrogen (some batteries may produce hydrogen when they fail) is detected to exceed the safe range, it indicates that there may be a potential fire risk caused by internal battery failure, thereby providing early warning for operation and maintenance personnel, so that they have enough time to take measures, such as checking the status of the battery pack, troubleshooting potential safety hazards, and preventing further development of fire accidents;

[0141] Continuously monitor the concentration of flammable gases in the environment to ensure that the air quality inside the energy storage power station is at a safe level; by promptly discovering and handling abnormal conditions such as flammable gas leaks, avoid the accumulation of flammable gases to dangerous concentrations that may cause explosions or fires, protect the safety of power station equipment and personnel, and reduce the economic losses and operational interruption risks caused by fire accidents to the energy storage power station;

[0142] Composite detectors are used to achieve comprehensive monitoring of multiple disasters and link fire protection systems;

[0143] Composite detectors are usually able to detect multiple danger signals at the same time, such as smoke, abnormal temperature rise, flames, etc. In a complex environment like an energy storage power station, the occurrence of a fire may be accompanied by the appearance of multiple characteristic signals. For example, in the early stage of battery thermal runaway, abnormal temperature rise may occur first, followed by smoke, and flames may appear when the fire develops to a certain extent. Composite detectors can comprehensively monitor these different types of danger signals to improve the accuracy and reliability of fire detection. Once any danger signal is detected to exceed the set threshold, the alarm mechanism will be triggered immediately. Compared with single-function detectors, it can more comprehensively respond to fire threats of different stages and types.

[0144] When the composite detector detects an abnormality and sends out an alarm signal, it can be linked with the fire protection system of the energy storage power station; for example, it can automatically start the fire sprinkler, start the ventilation system to exhaust smoke, shut down related electrical equipment, etc., to achieve rapid response and initial control of the fire, effectively curb the spread of the fire, reduce the degree of damage to the energy storage power station, buy time for subsequent fire rescue, and improve the fire safety performance of the entire energy storage power station.

[0145] Furthermore, the environmental parameters that the multi-parameter monitoring module 100 needs to collect include temperature, humidity and air pressure;

[0146] Among them, the battery pack will generate heat during the charging and discharging process. The ambient temperature and the battery's own temperature have a significant impact on the battery performance and safety. Too high a temperature may accelerate battery aging, reduce battery life, and even cause safety accidents such as thermal runaway. Too low a temperature may affect the battery's charging and discharging efficiency and capacity. By monitoring temperature parameters, heat dissipation or heating measures can be taken in time to ensure that the battery pack works within an appropriate temperature range and ensure the safe and stable operation of the energy storage power station.

[0147] Ambient humidity can affect the electrical insulation performance of equipment and the degree of corrosion of metal parts. A high humidity environment may cause condensation to form on the surface of electrical equipment, reducing insulation resistance and increasing the risk of leakage. At the same time, moisture combined with impurities in the air may form an electrolyte solution, accelerating the corrosion of metal parts and damaging equipment structures and connecting lines. Therefore, humidity monitoring helps to detect humidity anomalies in a timely manner, take dehumidification or moisture-proof measures, and protect equipment from damage in a humid environment.

[0148] Changes in air pressure reflect the stability of the environment and the operating status of the equipment to a certain extent. For example, in a sealed equipment compartment of an energy storage power station, abnormal air pressure may indicate that there is a leak in the equipment, affecting the gas environment and pressure balance inside the equipment. In addition, changes in air pressure may also be related to weather changes. Extreme changes in air pressure may affect the physical structure of the equipment, such as deformation of the outer shell. By monitoring air pressure parameters, potential problems can be discovered in advance to ensure the normal operation of the equipment.

[0149] In this embodiment, Figure 4 As shown, the operation process of the data processing unit 200 includes the following steps:

[0150] S2.1. Data collection: Collect the operating data of various equipment in the energy storage power station;

[0151] Data source: The data mainly comes from various sensors and monitoring equipment installed in energy storage power stations, such as temperature sensors, humidity sensors, voltage sensors, etc.

[0152] Data collection method: These real-time data are transmitted to the data processing unit through the data acquisition system (DCS) or distributed database connection.

[0153] S2.2. Data preprocessing: Clean, convert and standardize the collected raw data. Data preprocessing includes:

[0154] First, data cleaning is used to remove noise, outliers and duplicate data in the data to ensure the accuracy and consistency of the data; data cleaning includes two types: outlier processing and noise data smoothing;

[0155] (1) Outlier processing is based on the 3σ principle, and the specific algorithm is:

[0156] Assume that the monitoring data sequence is x1,x2,...,x n , first calculate the mean of the series and standard deviation σ;

[0157] Mean calculation formula:

[0158]

[0159] Standard deviation calculation formula:

[0160]

[0161] Then determine whether the data point is an outlier. If x i satisfy The data point is considered an outlier and can be processed using the following methods:

[0162] a. Delete outliers directly: i Delete from the data series to obtain a new data series for subsequent analysis;

[0163] b. Correct outliers: You can use the mean Or other reasonable values ​​(such as the average of adjacent data points, etc.) to replace x i , making the data series smoother and more reasonable.

[0164] (2) The moving average method is used for noise data smoothing. The specific algorithm is as follows:

[0165] For data sequences with noise, the moving average method is used for smoothing; let the window size be m (m is an odd number, and m <n);

[0166] For data points The smoothed result y i The calculation formula is:

[0167]

[0168] That is, in x i Take both before and after The data points are averaged as x iSmoothing value; for the data points at the beginning and end of the sequence, the window range can be adjusted appropriately according to the actual situation for smoothing, or other appropriate boundary processing methods can be used, such as using linear interpolation and other methods to supplement the boundary data before smoothing; the moving average method can effectively reduce the random noise in the data, making the data trend more obvious and facilitating subsequent analysis.

[0169] Second, data format conversion is used to convert data from different sources and formats into a unified format for subsequent processing and analysis; data format conversion includes analog to digital conversion (A / D conversion), conversion between different data types, and data encoding format conversion;

[0170] (1) The algorithm formula for analog-to-digital conversion (A / D conversion) (taking common linear conversion as an example) is:

[0171] Assume that the analog signal (such as voltage, current, etc.) collected by the sensor is V in , the reference voltage of the A / D converter is V ref , the resolution of the A / D converter is n bits (e.g. the common 12 bits, 16 bits, etc.);

[0172] The calculation formula for digital output D is:

[0173]

[0174] For example, for a 12-bit A / D converter, the reference voltage V ref =5V, when the input analog voltage V in =2,5V, according to the formula:

[0175]

[0176] The actual output will be processed according to the rounding rule of the A / D converter, which may be 2047 or 2048, etc.

[0177] This conversion converts continuous analog signals into discrete digital values ​​for easy processing and analysis by the data processing unit; different types of sensors may require different conversion circuits and formulas, but the basic principles are similar, which is to convert physical quantities into digital representations.

[0178] (2) The conversion formula between different data types (taking integer and floating point conversion as an example) is:

[0179] Convert integer to floating point:

[0180] Let the integer be I, to be converted to a single-precision floating point number (32 bits);

[0181] First, fill the binary representation of the integer to 32 bits (fill the high bits with 0), and then convert it according to the IEEE 754 floating point standard;

[0182] Sign bit: If I ≥ 0, the sign bit is 0; if I < 0, the sign bit is 1;

[0183] Exponent part: calculate the exponent offset value E = 127 + the number of binary bits of the integer - 1 (for non-zero integers); for example, for a 16-bit integer, its exponent offset value E = 127 + 16 - 1 = 142, convert E to binary representation and fill it into the exponent part (8 bits);

[0184] Mantissa: Remove the sign bit from the binary representation of the integer and fill it into the mantissa (23 bits). If the binary digits of the integer are less than 23 bits, fill the low digits with 0;

[0185] Convert floating point numbers to integers:

[0186] For single-precision floating-point numbers (32 bits), first parse the sign bit, exponent part, and mantissa part;

[0187] Calculate the actual exponent value e according to the exponent part (the binary value of the exponent part minus 127);

[0188] Calculate the actual value M of the mantissa (convert the binary representation of the mantissa part to decimal and add an implicit 1);

[0189] The value of integer I is I=(-1) 符号位 ×M×2 e , and then rounding is performed according to the integer representation range and precision requirements to obtain the final integer result; for example, if the calculation results in I=3.8, if it needs to be converted to an integer type (such as int), then according to the rounding rules, I=4 (rounded up) or I=3 (rounded down) may be obtained.

[0190] (3) The formula for data encoding format conversion (such as hexadecimal to decimal conversion) is:

[0191] Convert hexadecimal to decimal:

[0192] Let the hexadecimal number be H=h n h n-1 ...h i ...h1h0(h i It is a hexadecimal digit, and its value range is 0-9, AF);

[0193] The calculation formula for the decimal number D is:

[0194]

[0195] For example, the hexadecimal number 0x3A is converted to decimal: D = 3 × 16 1 +10×16 0 =48+10=58;

[0196] Convert decimal to hexadecimal:

[0197] Use the remainder method: divide the decimal number D by 16 to get the quotient Q and the remainder R. The remainder R is the lowest bit of the hexadecimal number. Then divide the quotient Q by 16 again, and repeat this process until the quotient is 0. Finally, arrange all the remainders from the back to the front to get the hexadecimal number.

[0198] For example, the decimal number 100 is converted to hexadecimal:

[0199] 100÷16=6......4 (the remainder 4 is the lowest digit in hexadecimal)

[0200] 6÷16=0......6 (the remainder 6 is the second lowest digit in hexadecimal)

[0201] So decimal 100 converts to hexadecimal 0x64;

[0202] In the actual energy storage power station monitoring data processing, these data format conversion formulas can be flexibly applied according to the specific sensor data type, communication protocol requirements and data processing needs to ensure that data can be accurately and efficiently transmitted and processed between different modules and systems.

[0203] Third, data standardization is used to scale or transform data according to specific specifications for easy comparison and analysis; data standardization includes minimum-maximum standardization, Z-score standardization, and decimal scaling standardization;

[0204] (1) The formula for Min-Max Scaling is:

[0205] Let the original data be x i ,i=1,2,...,n,x min is the minimum value in the original data, x max is the maximum value in the original data;

[0206] The standardized data y i The calculation formula is:

[0207]

[0208] After minimum-maximum normalization, the data is mapped to the [0,1] interval; for example, for a set of battery pack temperature data, the original temperature range is 0°C to 50°C, and a temperature value x i =25℃, then, xmin =0,x max =50, calculate the standardized value according to the formula This standardization method is suitable for situations where data distribution is relatively uniform and data needs to be mapped to a specific interval. It is commonly used in data visualization and other aspects.

[0209] (2) The formula for Z-mean Normalization is:

[0210] Let the original data be x i ,i=1,2,...,n, is the mean of the original data, σ is the standard deviation of the original data;

[0211] The standardized data z i The calculation formula is:

[0212]

[0213] After Z-score standardization, the mean of the data becomes 0 and the standard deviation becomes 1; for example, for a set of battery pack voltage data, the mean is calculated Standard deviation σ=0.1V, a certain voltage value x i =3.8V, then calculate the standardized value according to the formula Z-score standardization is applicable to situations where the data distribution is unknown and the data needs to have the characteristics of a standard normal distribution. It is widely used in data preprocessing of many machine learning algorithms because many algorithms assume that the data follows a normal distribution or works better under a normal distribution.

[0214] (3) The formula for decimal scaling is:

[0215] Let the original data be x i , j is the value that satisfies max(|x i |)<10j;

[0216] The calculation formula for the standardized data is:

[0217]

[0218] For example, for a set of battery capacity data, the maximum value is 1200Ah, then j = 4 (because 10 3 <1200<10 4 ), for a certain capacity value x i =800Ah, standardized value Decimal calibration standardization can effectively map the absolute value of data to a vicinity of an interval. It is suitable for data standardization scenarios where the data magnitude varies greatly but does not require strict normal distribution characteristics. It can improve computational efficiency and accuracy in some data mining and data analysis tasks.

[0219] In the processing of energy storage power station monitoring data, appropriate data standardization methods can be selected according to the characteristics of the data and subsequent analysis requirements. For example, if the main focus is on the relative size relationship of the data within a specific interval, the minimum-maximum standardization can be selected; if an algorithm based on the normal distribution assumption is to be applied later, Z-score standardization may be more suitable; and for situations where the data magnitude varies greatly and the distribution is not very regular, decimal calibration standardization may be a better choice. At the same time, in practical applications, it may be necessary to appropriately scale or offset the standardized data to meet specific business needs.

[0220] S2.3, Data storage: Store the preprocessed data in an appropriate location for subsequent use;

[0221] Storage location: Data is usually stored in a cloud database or a local database; cloud databases have higher availability and scalability, while local databases have higher security and controllability;

[0222] How it is stored: Data is stored in a structured or unstructured manner, depending on the type and purpose of the data.

[0223] S2.4, Data processing and analysis: Further processing and analysis of stored data, mainly including data processing, data analysis and data visualization;

[0224] Data processing: sorting, classifying, and summarizing data to facilitate subsequent analysis and mining;

[0225] Data analysis: Use statistics, data mining and other techniques to conduct in-depth analysis of data and discover patterns and trends in the data;

[0226] Data visualization: The analysis results are displayed in the form of charts, reports, etc., so that operation and maintenance personnel and management can quickly understand the operating status of the energy storage power station.

[0227] S2.5, Data application: Apply the processed and analyzed data to the actual operation of the energy storage power station;

[0228] Decision support: Provide decision support for the operation and scheduling, equipment maintenance, and fault handling of energy storage power stations;

[0229] System optimization: Based on the data analysis results, adjust the operation strategy of the energy storage power station, optimize equipment parameters, and improve system performance and reliability;

[0230] Early warning and alarm: Real-time monitoring of the operating status of the energy storage power station. Once an abnormality or failure is found, a warning or alarm signal will be issued in time so that the operation and maintenance personnel can take timely measures.

[0231] S2.6, Security and Maintenance: important links to ensure stable operation of data processing units and data security, including data security management, system maintenance and fault handling;

[0232] Data security: Take necessary security measures, such as encryption, backup, and permission management, to ensure the confidentiality, integrity, and availability of data;

[0233] System maintenance: Regularly maintain and upgrade the data processing unit, including hardware equipment maintenance, software system updates, etc., to ensure its stable operation and performance optimization;

[0234] Fault handling: Establish a fault handling mechanism so that when a data processing unit fails, the problem can be quickly located and effective repair measures can be taken.

[0235] In this embodiment, the intelligent algorithms built into the data processing unit (200) include at least a data anomaly detection algorithm, a fault prediction algorithm and an operation optimization algorithm; the data anomaly detection algorithm includes a threshold-based algorithm, a statistical analysis algorithm and a machine learning algorithm; the fault prediction algorithm includes a time series analysis algorithm and a neural network algorithm; the operation optimization algorithm includes a dynamic programming algorithm and a model predictive control algorithm.

[0236] Furthermore, the specific algorithm process of the data anomaly detection algorithm includes:

[0237] First, the threshold-based algorithm: set the normal threshold range of parameters such as voltage, current, temperature, humidity, air pressure, combustible gas content or smoke concentration. When the monitoring data exceeds this range, it is judged as abnormal; the algorithm formula is:

[0238] Let the monitoring parameter be x and the lower limit of the normal threshold be x min , the upper limit of the normal threshold is x max ;

[0239] If x <x min or x>x max , then the data x is judged to be abnormal; for example, in battery pack temperature monitoring, when the temperature exceeds 75°C, the battery pack is considered to be in an abnormal overheating state. This algorithm is simple and direct, easy to implement, and can quickly detect obvious abnormalities.

[0240] Second, statistical analysis algorithm: using the mean-standard deviation method, by calculating the mean and standard deviation of the monitoring data over a period of time, determine the distribution range of normal data; if the newly collected data deviates from this range to a certain extent, it is considered abnormal; it can take into account the statistical characteristics of the data and more flexibly adapt to data changes under different working conditions, but it has certain requirements for the stability and regularity of the data. The algorithm process is:

[0241] Assume that the monitoring data sequence is {x1,x2,...,x n};

[0242] Calculate the mean

[0243] Calculate standard deviation

[0244] Determine the interval of normal data as Where k is a coefficient set based on experience or actual conditions, usually 2 or 3;

[0245] if or The data is judged to be abnormal.

[0246] Third, machine learning algorithm: clustering algorithm is used to cluster the monitoring data under normal operation to form normal data clusters; when the new data point is far away from these normal clusters, it is judged as abnormal; clustering algorithm can automatically discover patterns in data and has good processing capabilities for complex nonlinear relationships, but the computational complexity is relatively high and requires a certain amount of training data. The K-Means algorithm is used, and its algorithm process is as follows:

[0247] Assume that the monitoring data is an n-dimensional vector set X = {x1, x2, ..., x m}, where x i ={x i1 ,x i2 ,...,x in};

[0248] Randomly select K initial cluster centers C = {c1, c2, ..., c K}, where c j ={c j1 ,c j2 ,...,c jn}, j = 1, 2, ..., K;

[0249] For each data point x i , calculate it to each cluster center c j The distance d ij , usually using the Euclidean distance formula

[0250] The data point x i Assign to the cluster to which the nearest cluster center belongs;

[0251] Recalculate the cluster center of each cluster, that is, for the jth cluster, the new cluster center c j The value of each dimension of is the average value of the corresponding dimension of all data points in the cluster, that is, Where N j is the number of data points in the jth cluster;

[0252] Repeat the above steps of assigning data points and updating cluster centers until the cluster centers no longer change significantly (converge);

[0253] For a new data point x new , calculate its distance to each cluster center, if its distance to the nearest cluster center is greater than a certain set threshold, then determine x new is abnormal.

[0254] Furthermore, the specific algorithm process of the fault prediction algorithm includes:

[0255] First, the time series analysis algorithm: Use the ARIMA model to model and analyze historical monitoring data (such as the time series of the voltage and current of the battery pack) to predict the trend of future data; if the predicted value deviates greatly from the actual monitoring value, or the predicted trend shows that abnormal changes are about to occur (such as the continuous downward trend of voltage may indicate that the battery capacity is declining or is about to fail), then an early warning of the fault will be issued; time series analysis is suitable for data with time correlation, and can use historical information to predict future trends, but it has certain assumptions about the stability of the data. The specific algorithm process includes:

[0256] Model building process:

[0257] Assume that the monitoring data sequence is {y t}, t = 1, 2, ..., T;

[0258] First, test the data for stationarity. Common test methods include ADF (Augmented Dickey-Fuller) test. If the data is non-stationary, differential processing is required until the data is stable. Let the number of differentials be d;

[0259] Determine the autoregressive order p and moving average order q of the ARIMA model; the values ​​of p and q can be preliminarily determined by observing the graphs of the autocorrelation function (ACF) and partial autocorrelation function (PACF), or the information criteria (such as AIC, BIC, etc.) can be used to select the optimal combination of p and q;

[0260] The formula for establishing the ARIMA (p, d, q) model is:

[0261]

[0262] in is an autoregressive operator, θ(B)=1+θ1B+θ2B 2 +...+θ q B q is the moving average operator, and B is the lag operator (i.e., B k y t =y t -k),∈ t is a white noise sequence;

[0263] Prediction process:

[0264] Estimate the parameters of the ARIMA model and θ1,θ2,...,θ q , we can use methods such as least squares or maximum likelihood estimation;

[0265] The predicted value for the future time T+h, (h=1,2,...) It can be obtained by recursive calculation. For example, for the ARIMA (1,1,1) model, the prediction formula is:

[0266]

[0267] If the predicted value Compared with the actual monitoring value y T+h If the deviation is large (which can be determined by setting an error threshold), or if the forecast trend shows that an abnormal change is about to occur (such as the forecast value continues to decrease or increase by more than a certain amount), an early warning of the fault will be issued.

[0268] Second, neural network algorithm: using recurrent neural network RNN ​​and its variant LSTM, through learning a large amount of historical data, to establish a complex nonlinear relationship model between the battery pack operating status and various monitoring parameters; neural network can capture long-term dependencies and hidden patterns in the data, thereby predicting the probability of faults; for example, based on the battery pack's past charge and discharge current, voltage, and temperature changes, predict whether there may be internal battery short circuits and other faults in the future. Neural network algorithms have strong learning capabilities, but the training process may require a lot of computing resources and time, and the model's interpretability is relatively weak. The specific algorithm process includes:

[0269] RNN basic structure and algorithm process:

[0270] The basic structure of RNN includes input layer, hidden layer and output layer; let the input sequence be {x t}, t=1,2,...,T, the hidden layer state is {ht}, the output is {y t};

[0271] The update formula of the hidden layer state is:

[0272] h t =f(Ux t +Wh t-1 +b)

[0273] Where U is the weight matrix from the input layer to the hidden layer, W is the weight matrix from the hidden layer to the hidden layer (self-circular connection), b is the bias vector of the hidden layer, and f(·) is the activation function, such as tanh or ReLU;

[0274] The output formula of the output layer is:

[0275] y t =g(Vh t +c)

[0276] Where V is the weight matrix from the hidden layer to the output layer, c is the bias vector of the output layer, and g(·) is the activation function, such as softmax.

[0277] During training, the back propagation algorithm (BPTT-Back Propagation Through Time) is used to update the weight matrices U, W, and V and the bias vectors b and c; the goal is to minimize the loss function, such as the mean square error (MSE) loss function:

[0278]

[0279] in is the true output value;

[0280] Use LSTM to improve RNN, LSTM structure and algorithm process:

[0281] LSTM improves the unit structure of the hidden layer based on RNN; the LSTM unit contains input gate, forget gate and output gate;

[0282] The formula of the forget gate is:

[0283] f t =σ(W f x t +U f h t-1 +b f )

[0284] Where W f , U f is the corresponding weight matrix, b fis the bias vector, σ(·) is the sigmoid function;

[0285] The formula for the input gate is:

[0286] i t =σ(W i x t +U i h t-1 +b i )

[0287] And there is a new candidate state

[0288] The formula for the output gate is:

[0289] o t =σ(W o x t +U o h t-1 +b o )

[0290] The update formula of the hidden layer state is:

[0291]

[0292] The output formula of the output layer is similar to that of RNN;

[0293] After training the LSTM model with a large amount of historical data, the new monitoring data is input into the model to obtain a predicted output of the probability of fault occurrence; if the predicted probability exceeds the set threshold, a fault warning is issued.

[0294] Furthermore, the specific algorithm process of running the optimization algorithm includes:

[0295] First, dynamic programming algorithm: Taking into account the operating constraints of the energy storage power station (such as battery pack charging and discharging power limits, energy storage capacity limits, etc.) and goals (such as minimizing operating costs, maximizing energy utilization efficiency, etc.), the charging and discharging strategies of the battery pack are optimized through dynamic programming methods. For example, according to the electricity prices in different time periods, the forecast of renewable energy power generation and load demand, determine whether the battery pack is charging, discharging or on standby at each moment to maximize the economic benefits of the energy storage power station. The dynamic programming algorithm can handle multi-stage decision-making problems and find the global optimal solution, but the computational complexity increases with the size of the problem. The algorithm process includes:

[0296] Assume that the operation cycle of the energy storage power station is divided into T time periods, each of which is represented by t=1, 2, ..., T;

[0297] Define the state variable S tIndicates the status of the energy storage power station at the beginning of time period t, including information such as the remaining power of the battery pack;

[0298] Define the decision variable x t represents the decision taken at time period t, for example, the battery pack is charged (x t =1), discharge (x t = -1) or standby (x t =0);

[0299] Define the cost function C(S t ,x t ) indicates that in state S t Take decision x t The costs incurred, such as charging costs, discharge loss costs, etc.;

[0300] Define the transfer function S t+1 =T(S t ,x t ) indicates that according to the current state S t and decision x t , get the state at the beginning of the next time period t+1;

[0301] The dynamic programming algorithm solves the optimal decision sequence by following these steps:

[0302] Initialization: For t = T, calculate the terminal cost That is, find the decision that minimizes the cost in the last time period;

[0303] Recursive calculation: For t = T-1, T-2, ..., 1, calculate That is, considering the current cost and the optimal cost of the next time period, find the decision that minimizes the total cost in the current time period;

[0304] Optimal decision sequence: By backtracking the above calculation process, we can get the optimal decision sequence from t = 1 to t = T This sequence is the charging and discharging strategy that minimizes the cost of the energy storage power station throughout the entire operating cycle.

[0305] Second, the model predictive control algorithm (MPC): Based on the dynamic model of the system (such as the charging and discharging model of the energy storage battery pack, the interaction model with the power grid or other energy systems, etc.), in each control cycle, according to the current monitoring data and the predicted future system state (such as wind power generation power and load demand changes in the future period of time), the optimal control input (such as the charging and discharging power instructions of the battery pack) is calculated. The MPC algorithm can optimize the system operation in real time and has good adaptability to dynamically changing environmental and system conditions, but it requires an accurate system model and a relatively fast calculation speed to ensure real-time performance; the algorithm process includes:

[0306] Assume that the dynamic model of the energy storage power station can be expressed by the following discrete time state space equation:

[0307] x k+1 =f(x k ,u k )+w k

[0308] y k =g(x k )+v k

[0309] where x k is the state vector of the system at time k (such as the battery pack’s power, temperature, and other state information), u k is the control input vector (e.g., the charge and discharge power command of the battery pack), y k is the measured output vector (e.g., the actual measured battery pack charge, temperature, etc.), f(·) and g(·) are the dynamic functions of the system, and w k and v k are process noise and measurement noise respectively;

[0310] In each control cycle k, the MPC algorithm performs the following steps:

[0311] Prediction model: Based on the state x at the current time k k and the measured output y k , and the predicted future system state (such as wind power generation and load demand changes in the future), using the system's dynamic model to predict the system state in the next N time steps and measurement output in represents the state at time k+i predicted at time k, represents the measured output at time k+i predicted at time k;

[0312] Objective function setting: Define the objective function Where L(·) is the loss function that measures the system performance (e.g., losses related to battery pack charging and discharging efficiency, grid balance, etc.), and M(·) is the function that measures the control input cost (e.g., costs related to battery pack charging and discharging power limits, energy consumption, etc.);

[0313] Optimization solution: By solving the minimum value of the objective function J, the optimal control input sequence for the next N-1 time steps is obtained. That is, to find the charging and discharging power instruction sequence that optimizes system performance and control cost in the future;

[0314] Implement control: Input the first optimal control Apply to the system and then repeat the above steps in the next control cycle.

[0315] In addition, the energy storage power station monitoring device of this embodiment further includes a structural design module 500, which adopts a modular anti-seismic, dust-proof, and waterproof design to ensure the stable operation of the device in a complex environment; including:

[0316] (1) Optimization of seismic design:

[0317] (1.1) Use seismic isolation technology: Install seismic isolation pads or supports, such as rubber seismic isolation pads or spring seismic isolation supports, between the monitoring device and the installation foundation. These seismic isolation elements can effectively absorb and dissipate vibration energy and reduce the impact of vibration waves on the monitoring device. When vibration occurs, the seismic isolation pad can cushion the movement of the equipment through its own deformation, preventing the equipment from damaging internal components or increasing sensor measurement errors due to severe shaking.

[0318] (1.2) Strengthen the equipment frame structure: Use high-strength metal materials (such as aluminum alloy or stainless steel) to make the equipment frame, and optimize the frame structure design. Increase the wall thickness and reinforcement ribs of the frame to improve its overall rigidity and load-bearing capacity. Use finite element analysis and other methods to simulate and analyze the frame structure, optimize its force distribution, and ensure that the frame can remain stable under external forces such as severe vibration, without serious deformation or fracture, so as to protect the internal monitoring modules and circuits and other components.

[0319] (1.3) Install anti-seismic dampers: Install anti-seismic dampers, such as hydraulic dampers or friction dampers, at appropriate locations inside or outside the equipment. The damper can generate damping force when the equipment vibrates, suppress the vibration amplitude, and restore the equipment to a stable state more quickly. For example, if a damper is installed on the mounting bracket of the battery pack, when the battery pack shakes during severe vibration, the damper can effectively control its swing amplitude, reduce the risk of collision between battery packs, and also help protect the cables connecting the battery pack and the sensor cables from excessive stretching or damage.

[0320] (2) Dust-proof design optimization:

[0321] (2.1) Improvement of sealing design: The outer shell of the monitoring device is sealed and high-quality sealing materials (such as rubber sealing rings or silicone sealing strips) are used to ensure the sealing of the outer shell joints. Sealing measures are added to the cabinet doors, interfaces and other parts of the equipment to prevent dust from entering the equipment through the gaps. For example, a double-layer sealing structure is designed on the edge of the cabinet door to improve the sealing effect; for cable interfaces, a sealing rubber sleeve or sealing joint is used to ensure that dust is not introduced when the cable passes through the outer shell.

[0322] (2.2) Air filtration system upgrade: Install a high-efficiency air filter (HEPA filter) at the vent or air inlet of the equipment to effectively filter tiny dust particles in the air (the filtration efficiency can reach more than 99.97%, and can filter particles with a particle size of 0.3 microns and above). Replace the air filter regularly to ensure its filtration effect. At the same time, the air filtration system can be designed with a self-cleaning function, such as removing dust on the filter surface through reverse airflow blowing or vibration, to extend the service life of the filter and reduce maintenance workload.

[0323] (2.3) Optimize the internal layout to prevent dust accumulation: Rationally design the internal layout of the equipment to avoid dead corners where dust can easily accumulate. For example, install key components such as circuit boards and sensors at a position higher than the bottom of the equipment and maintain a certain distance to prevent dust from accumulating at the bottom due to gravity and affecting the heat dissipation and normal operation of the equipment. For some components that are prone to static electricity and dust absorption, use anti-static materials for surface treatment or install static electricity elimination devices to reduce dust absorption.

[0324] (3) Waterproof design optimization:

[0325] (3.1) Strengthen the waterproof design of the shell: Choose shell materials with good waterproof performance, such as waterproof plastic or metal shell, and apply waterproof coating. The design of the shell should meet the IP protection level standard, such as reaching IP65 or higher, to ensure that the equipment can still maintain waterproof performance under certain water pressure and water flow impact. The connection parts of the shell adopt a sealed waterproof structure, such as using threaded connection with sealant, or using a waterproof buckle design to ensure that the connection is leak-proof.

[0326] (3.2) Waterproof and breathable valve installation: Install a waterproof and breathable valve on the device casing to balance the air pressure inside and outside the device and prevent moisture from being inhaled due to negative pressure inside the device caused by temperature changes. The waterproof and breathable valve has the dual functions of being waterproof and breathable, allowing gas to freely enter and exit the device while preventing moisture from entering. For example, installing a waterproof and breathable valve on the battery pack casing can effectively prevent the battery pack from inhaling humid air due to internal gas expansion or contraction during the charging and discharging process, thereby avoiding battery short circuits and other faults.

[0327] (3.3) Waterproof cable and interface treatment: Use waterproof cables and perform special waterproof treatment on the interfaces between cables and equipment. The cable interfaces use waterproof plugs and sockets and are sealed after connection, such as wrapping with waterproof tape or injecting sealant. For the parts of the cables exposed to the outside, waterproof casings are used for protection to prevent the cables from being damaged due to long-term exposure to humid environments, which affects signal transmission and equipment power supply. In addition, waterproof isolation or protection treatment is performed on the components inside the device that may be exposed to moisture, such as waterproof coating of circuit boards and waterproof packaging of sensitive parts of sensors.

[0328] (4) Modular design optimization:

[0329] (4.1) Functional module independence and standardization: The monitoring device is divided into multiple independent functional modules, such as multi-parameter monitoring module 100, data processing module 200, communication module 300, display and alarm module 400, etc. Each module has clear functions and interface standards. Standardized connection methods are used between modules, such as plug-in connectors or bus interfaces, to facilitate the rapid installation, removal and replacement of modules. In this way, when a module fails, it can be easily repaired or replaced separately without affecting the normal operation of other modules, thereby improving the maintainability and reliability of the equipment.

[0330] (4.2) Module universal design: Design universal modules so that they can be applied to energy storage power stations of different specifications and types. For example, the multi-parameter monitoring module 100 can be flexibly configured according to the voltage, current range and sensor type of different battery packs, and can be adapted to a variety of application scenarios through software settings or replacement of a small number of hardware components. This universal design can reduce the customization cost of equipment, improve the reusability of modules, and facilitate large-scale production and inventory management.

[0331] (4.3) Application of module hot-swap technology: Hot-swap technology is used to enable modules to be plugged in and out while the equipment is running. This requires corresponding support in hardware and software design, such as adding hot-swap protection circuits in circuit design and implementing real-time monitoring and dynamic configuration of module status in software. When a module needs to be replaced, there is no need to stop the operation of the entire monitoring device, which greatly improves the system availability and maintenance efficiency and reduces the downtime of the energy storage power station caused by equipment maintenance.

[0332] (5) Heat dissipation design integration:

[0333] (5.1) Application of high-efficiency heat dissipation materials: Use high-efficiency heat dissipation materials, such as thermally conductive silicone sheets or graphite heat sinks, between the heat-generating components inside the monitoring device (such as the processor chip and power amplifier in the data processing unit) and the outer shell. These materials have good thermal conductivity and can quickly and evenly transfer heat to the outer shell, and exchange heat with the external environment through the outer shell. For example, a thermally conductive silicone sheet is attached to the surface of the processor chip to make it in close contact with the metal shell, thereby improving the heat dissipation efficiency and preventing the chip from being degraded or damaged due to overheating.

[0334] (5.2) Optimization of heat dissipation structure: Design a reasonable heat dissipation structure, such as adding heat dissipation fins or heat dissipation holes on the outer shell to increase the heat dissipation area and promote air convection. For some high-power equipment, fan forced air cooling can be used to automatically adjust the fan speed according to the internal temperature of the equipment to ensure effective heat dissipation under different load conditions. At the same time, optimize the heat dissipation channel to ensure that cold air can flow smoothly through the heat-generating components, take away the heat, and the hot air can be discharged from the equipment in time. For example, a reasonable air duct is designed inside the equipment to allow air to enter from the low-temperature area and be discharged from the high-temperature area after passing through the heat-generating components, forming an effective heat dissipation cycle.

[0335] (5.3) Temperature monitoring and control integration: Integrate temperature monitoring sensors in the structural design module to monitor the temperature of key parts inside the equipment in real time. When the temperature exceeds the set threshold, the control circuit automatically adjusts the heat dissipation measures, such as increasing the fan speed or issuing an alarm signal. In addition, it can also be combined with the overall temperature control system of the energy storage power station to intelligently control the operation of the heat dissipation equipment according to the ambient temperature and equipment operating status to achieve a balance between energy saving and efficient heat dissipation. For example, in a low temperature environment, appropriately reduce the heat dissipation intensity to reduce energy consumption; in a high temperature environment or when the equipment is running at high load, enhance the heat dissipation capacity to ensure the normal operation of the equipment.

[0336] Application Example 1

[0337] In order to realize real-time monitoring and alarm of the operating status of the battery pack, the energy storage power station intelligent monitoring device of Example 1 is installed in an energy storage power station with a capacity of 500kWh. The specific implementation method is as follows:

[0338] Operation mechanism: The multi-parameter monitoring module 100 collects the voltage, current, temperature and environmental parameters (such as humidity, air pressure, etc.) of the battery pack in real time, with a sampling frequency of once per second to ensure the timeliness and accuracy of the data.

[0339] After receiving the monitoring data, the data processing unit 200 uses a built-in algorithm to analyze the operating status of the battery pack; when it is detected that the temperature of a battery pack exceeds 75°C, the system determines it as an overheating state and triggers the abnormality handling process.

[0340] Abnormal processing: The display and alarm unit 400 displays the number and temperature parameters of the currently overheated battery pack on the LCD screen, and activates the sound and light alarm to remind the on-site operation and maintenance personnel;

[0341] The communication module 300 sends the abnormal information to the management platform of the operation and maintenance center through the wireless network, including the time when the abnormality occurred, the battery pack number and the real-time parameters, so as to facilitate remote control.

[0342] Effect verification: Through the real-time monitoring and alarm function of the device, operation and maintenance personnel can quickly respond to problems at the early stage and take cooling measures to prevent the battery pack from being damaged due to continuous overheating, thereby improving the safety and service life of the battery pack.

[0343] Application Example 2

[0344] In order to realize intelligent operation optimization of energy storage power station, in an energy storage system supporting wind power generation, the energy storage power station monitoring device of Example 1 is deployed to realize dynamic optimization of the energy storage system, which is specifically implemented as follows:

[0345] Operation mechanism: The multi-parameter monitoring module 100 continuously collects the battery charging and discharging data of the energy storage power station, the ambient temperature and the real-time power output of the wind power generation.

[0346] The data processing unit 200 dynamically calculates the optimal charging and discharging strategy based on the collected data, combined with the volatility of wind power generation and the health status of the energy storage battery group. For example, when the wind power output is too high and exceeds the grid demand, the system will prioritize the charging mode of the energy storage battery; and when the wind power is insufficient, the energy storage battery will be released to balance the grid load.

[0347] Communication and feedback: Through the communication module 300, the device synchronizes the calculated charging and discharging strategy to the wind power generation control center to achieve coordinated control of the energy storage system and the power generation system.

[0348] At the same time, operation and maintenance personnel can view the operating status, charging and discharging efficiency, and overall energy utilization of the battery pack through the remote management platform.

[0349] Effect verification: Through the intelligent optimization function of the device, the energy storage power station realizes dynamic adjustment of charging and discharging efficiency, significantly improves the wind power absorption capacity, reduces overcharging or over-discharging of the battery pack, extends the service life of the equipment, and improves the overall energy utilization efficiency by about 15%.

[0350] Those skilled in the art will appreciate that the process of implementing all or part of the steps of the above embodiments may be completed by hardware, or may be completed by instructing related hardware through a program.

[0351] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A monitoring device for an energy storage power station, characterized in that: It comprises a multi-parameter monitoring module (100), a data processing unit (200), a communication module (300) and a display and alarm unit (400) which are communicatively connected in sequence; wherein: The multi-parameter monitoring module (100) collects the voltage, current and environmental parameters of the battery pack by integrating a high-sensitivity sensor array; The data processing unit (200) has a built-in intelligent algorithm, which can quickly identify abnormal conditions in operation by analyzing monitoring data in real time, and generate an optimized operation plan in combination with historical data; The communication module (300) supports wireless and wired transmission, and is connected to a cloud management platform, so that operation and maintenance personnel can remotely view the device status and receive abnormal alarms; The display and alarm unit (400) displays key parameters via an LCD screen and is also equipped with an audible and visual alarm function to provide an immediate warning when an abnormal situation occurs.

2. The energy storage power station monitoring device according to claim 1, characterized in that: In the multi-parameter monitoring module (100), the high-sensitivity sensor array comprises at least a voltage sensor, a current sensor, a temperature sensor, an ambient humidity sensor, an air pressure sensor, a combustible gas sensor and a composite detector.

3. The energy storage power station monitoring device according to claim 1, characterized in that: The environmental parameters that need to be collected by the multi-parameter monitoring module (100) include temperature, humidity and air pressure.

4. The energy storage power station monitoring device according to claim 1, characterized in that: The operation process of the data processing unit (200) comprises the following steps: S2.

1. Data collection: Collect the operating data of various equipment in the energy storage power station; S2.2, Data preprocessing: Cleaning, format conversion and standardization of the collected raw data; S2.3, Data storage: Store the preprocessed data in an appropriate location for subsequent use; S2.4, Data processing and analysis: Further processing and analysis of stored data, mainly including data processing, data analysis and data visualization; S2.5, Data application: Apply the processed and analyzed data to the actual operation of the energy storage power station; S2.6, Security and Maintenance: An important link to ensure the stable operation of the data processing unit and protect data security, including data security management, system maintenance and troubleshooting.

5. The energy storage power station monitoring device according to claim 4, characterized in that: In step S2.2, preprocessing the data includes: Data cleaning is used to remove noise, outliers and duplicate data in the data to ensure the accuracy and consistency of the data. Data cleaning includes two types: outlier processing and noise data smoothing. Outlier processing is based on the 3σ principle, and the specific algorithm is: Assume that the monitoring data sequence is x1,x2,...,x n , first calculate the mean of the series and standard deviation σ; Mean calculation formula: Standard deviation calculation formula: Then determine whether the data point is an outlier. If x i satisfy The data point is considered an outlier and can be processed using the following methods: a. Delete outliers directly: i Delete from the data series to obtain a new data series for subsequent analysis; b. Correct outliers: You can use the mean To replace x i , making the data series smoother and more reasonable; The moving average method is used for noise data smoothing. The specific algorithm is as follows: For data sequences with noise, the moving average method is used for smoothing; let the window size be m (m is an odd number, and m <n); For a data point x i ( and ), the smoothed result y i The calculation formula is: That is, in x i Take both before and after The data points are averaged as x i The smoothing value of Data format conversion is used to convert data from different sources and formats into a unified format; data format conversion includes analog to digital conversion, conversion between different data types, and data encoding format conversion; Data normalization is used to scale or transform data according to specific specifications for easy comparison and analysis; data normalization includes minimum-maximum normalization, Z-score normalization, and decimal scaling normalization.

6. The energy storage power station monitoring device according to claim 1, characterized in that: The intelligent algorithm built into the data processing unit (200) includes at least a data anomaly detection algorithm, a fault prediction algorithm and an operation optimization algorithm; Data anomaly detection algorithms include threshold-based algorithms, statistical analysis algorithms, and machine learning algorithms; Fault prediction algorithms include time series analysis algorithms and neural network algorithms; The optimization algorithms run include dynamic programming algorithms and model predictive control algorithms.

7. The energy storage power station monitoring device according to claim 6, characterized in that: The specific algorithm process of the data anomaly detection algorithm includes: First, the threshold-based algorithm: set the normal threshold range of voltage, current, temperature, humidity, air pressure, combustible gas content or smoke concentration parameters. When the monitoring data exceeds the range, it is judged as abnormal. Second, statistical analysis algorithm: using the mean-standard deviation method, by calculating the mean and standard deviation of the monitoring data over a period of time, the distribution range of normal data is determined; if the newly collected data deviates from this range to a certain extent, it is considered abnormal; Third, machine learning algorithm: clustering algorithm is used to cluster the monitoring data under normal operation to form normal data clusters; when the new data point is far away from these normal clusters, it is judged as abnormal; K-Means algorithm is used, and its algorithm process is as follows: Assume that the monitoring data is an n-dimensional vector set X = {x1, x2, ..., x m }, where x i ={x i1 ,x i2 ,...,x in }; Randomly select K initial cluster centers C = {c1, c2, ..., c K }, where c j ={c j1 ,c j2 ,...,c jn }, j = 1, 2, ..., K; For each data point x i , calculate it to each cluster center c j The distance d ij , usually using the Euclidean distance formula The data point x i Assign to the cluster to which the nearest cluster center belongs; Recalculate the cluster center of each cluster, that is, for the jth cluster, the new cluster center c j The value of each dimension of is the average value of the corresponding dimension of all data points in the cluster, that is, Where N j is the number of data points in the jth cluster; Repeat the above steps of assigning data points and updating cluster centers until the cluster centers no longer change significantly; For a new data point x new , calculate its distance to each cluster center, if its distance to the nearest cluster center is greater than a certain set threshold, then determine x new is abnormal.

8. The energy storage power station monitoring device according to claim 6, characterized in that: The specific algorithm process of the fault prediction algorithm includes: First, the time series analysis algorithm: Use the ARIMA model to model and analyze historical monitoring data to predict the trend of future data; if the predicted value deviates greatly from the actual monitoring value, or the predicted trend shows that abnormal changes are about to occur, an early warning fault will be issued; the specific algorithm process includes: Model building process: Assume that the monitoring data sequence is {y t }, t = 1, 2, ..., T; First, test the data for stationarity; if the data is non-stationary, differential processing is required until the data is stable; let the number of differentials be d; Determine the autoregressive order p and moving average order q of the ARIMA model; The formula for establishing the ARIMA (p, d, q) model is: in is an autoregressive operator, θ(B)=1+θ1B+θ2N 2 +...+θ q B q is the moving average operator, and B is the lag operator (i.e., B k y t =y t -k),∈ t is a white noise sequence; Prediction process: Estimate the parameters of the ARIMA model and θ1,θ2,...,θ q , the least squares method or maximum likelihood estimation method can be used; The predicted value for the future time T+h, (h=1,2,...) It can be obtained by recursive calculation; If the predicted value Compared with the actual monitoring value y T+h If the deviation is large, or the forecast trend shows that an abnormal change is about to occur, an early warning of the failure will be given; Second, neural network algorithm: using recurrent neural network RNN ​​and its variant LSTM, by learning a large amount of historical data, a complex nonlinear relationship model between the battery pack operating status and various monitoring parameters is established; the specific algorithm process includes: RNN basic structure and algorithm process: The basic structure of RNN includes input layer, hidden layer and output layer; let the input sequence be {x t }, t=1,2,...,T, the hidden layer state is {h t }, the output is {y t }; The update formula of the hidden layer state is: h t =f(Ux t +Wh t-1 +b) Where U is the weight matrix from the input layer to the hidden layer, W is the weight matrix from the hidden layer to the hidden layer, b is the bias vector of the hidden layer, and f(·) is the activation function; The output formula of the output layer is: y t =g(Vh t +c) Where V is the weight matrix from the hidden layer to the output layer, c is the bias vector of the output layer, and g(·) is the activation function; During the training process, the back propagation algorithm is used to update the weight matrices U, W, and V and the bias vectors b and c; the goal is to minimize the loss function, using the mean square error loss function: in is the true output value; Use LSTM to improve RNN, LSTM structure and algorithm process: The LSTM unit contains an input gate, a forget gate, and an output gate; The formula of the forget gate is: f t =σ(W f x t +U f h t-1 +b f ) Where W f , U f is the corresponding weight matrix, b f is the bias vector, σ(·) is the sigmoid function; The formula for the input gate is: i t =σ(W i x t +U i h t-1 +b i ) And there is a new candidate state The formula for the output gate is: the t =σ(W o x t +U o h t-1 +b o ) The update formula of the hidden layer state is: The output formula of the output layer is similar to that of RNN; After training the LSTM model with a large amount of historical data, the new monitoring data is input into the model to obtain a predicted output of the probability of fault occurrence; if the predicted probability exceeds the set threshold, a fault warning is issued.

9. The energy storage power station monitoring device according to claim 6, characterized in that: The specific algorithm process of running the optimization algorithm includes: First, dynamic programming algorithm: Under the condition of considering the operation constraints and goals of the energy storage power station, the charging and discharging strategy of the battery pack is optimized through dynamic programming method; the algorithm process includes: Assume that the operation cycle of the energy storage power station is divided into T time periods, each of which is represented by t=1, 2, ..., T; Define the state variable S t Indicates the state of the energy storage power station at the beginning of time period t, including the remaining power information of the battery pack; Define the decision variable x t represents the decision taken in time period t; Define the cost function C(S t ,x t ) indicates that in state S t Take decision x t the costs incurred; Define the transfer function S t+1 =T(S t ,x t ) indicates that according to the current state S t and decision x t , get the state at the beginning of the next time period t+1; The dynamic programming algorithm solves the optimal decision sequence by following these steps: Initialization: For t = T, calculate the terminal cost That is, find the decision that minimizes the cost in the last time period; Recursive calculation: For t = T-1, T-2, ..., 1, calculate That is, considering the current cost and the optimal cost of the next time period, find the decision that minimizes the total cost in the current time period; Optimal decision sequence: By backtracking the above calculation process, we can get the optimal decision sequence from t = 1 to t = T This sequence is the charging and discharging strategy that minimizes the cost of the energy storage power station during the entire operation cycle; Second, the model predictive control algorithm: Based on the dynamic model of the system, in each control cycle, the optimal control input is calculated according to the current monitoring data and the predicted future system state; the algorithm process includes: Assume that the dynamic model of the energy storage power station can be expressed by the following discrete time state space equation: x k+1 =f(x k ,u k )+w k y k =g(x k )+v k where x k is the state vector of the system at time k, u k is the control input vector, y k is the measured output vector, f(·) and g(·) are the dynamic functions of the system, and w k and v k are process noise and measurement noise respectively; In each control cycle k, the MPC algorithm performs the following steps: Prediction model: Based on the state x at the current time k k and the measured output y k , and the predicted future system state, using the system's dynamic model to predict the system state for N time steps in the future and measurement output in represents the state at time k+i predicted at time k, represents the measured output at time k+i predicted at time k; Objective function setting: Define the objective function Where L(·) is the loss function that measures the system performance, and M(·) is the function that measures the control input cost; Optimization solution: By solving the minimum value of the objective function J, the optimal control input sequence for the next N-1 time steps is obtained. That is, to find the charging and discharging power instruction sequence that optimizes system performance and control cost in the future; Implement control: Input the first optimal control Apply to the system and then repeat the above steps in the next control cycle.

10. The energy storage power station monitoring device according to claim 1, characterized in that: It also includes a structural design module (500), which adopts a modular anti-vibration, dust-proof and waterproof design to ensure the stable operation of the device in a complex environment.

Images