Ship diesel / light / energy storage hybrid power system load prediction and dynamic power matching method

Abstract

The present application relates to the technical field of prediction and matching, in particular to a ship diesel / light / energy storage hybrid power system load prediction and dynamic power matching method, comprising multi-source data acquisition and preprocessing: real-time data acquisition, establishment of a unified timestamp time series dataset; multi-modal data fusion processing: abnormal value detection and missing data reconstruction, generation of feature fusion dataset; dynamic load prediction modeling: construction of a hybrid prediction model combining deep learning and time series analysis; dynamic power distribution scheme generation: generation of a dynamic power distribution scheme according to the optimal power distribution ratio; real-time dynamic adjustment and feedback optimization: based on the deviation value of the power system real-time operation data and the power distribution scheme, triggering the dynamic adjustment mechanism to update the power distribution parameters; control instruction generation and execution: converting the updated power distribution parameters into a control instruction set for each subsystem. The present application significantly enhances the system operation stability, economy and intelligent level.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of prediction and matching, in particular to a ship diesel / light / energy storage hybrid power system load prediction and dynamic power matching method. BACKGROUND

[0002] With the development of new energy and green shipping concept, diesel / light / energy storage hybrid power system is increasingly widely used in intelligent ships. Such system is usually composed of diesel generator set, photovoltaic power generation unit and energy storage battery, and meets the dynamic load demand of the ship under different navigation conditions through multi-source joint energy supply. However, the existing system mainly relies on static setting or rule-driven power distribution strategy, lacks fine response mechanism to real-time load fluctuation, equipment aging state and energy storage SOC (state of charge), resulting in low power distribution efficiency, unbalanced energy utilization, and even may cause power overload, energy storage over-discharge and other problems. In addition, the traditional control architecture is difficult to realize dynamic coordination between multi-energy subsystems, and the system response is lagging and the control granularity is coarse, which has been difficult to meet the demand of intelligent ships for high reliability and high economic efficiency of power system.

[0003] In order to solve the above problems, it is necessary to build a diesel / light / energy storage system power management method which integrates multi-source data perception, prediction-driven optimization and distributed execution control. Real-time acquisition of multi-modal data including power load, light intensity, SOC, diesel engine operating parameters and ship navigation state, construction of load prediction model combining deep learning and time series analysis; according to the predicted load, equipment state and aging factor, the optimal power distribution ratio is calculated, and the dynamic correction is carried out by combining the energy storage SOC limit; through the feedback optimization mechanism, the distribution result is adjusted in real time, and the updated power instruction is converted into executable control quantity, realizing fine collaborative control between each subsystem, so as to improve the energy response accuracy, ensure the safe and stable operation of the system, and significantly improve the intelligent level of the ship hybrid power system. SUMMARY

[0004] Based on the above purpose, the present application provides a ship diesel / light / energy storage hybrid power system load prediction and dynamic power matching method.

[0005] The ship diesel / light / energy storage hybrid power system load prediction and dynamic power matching method comprises the following steps:

[0006] S1, multi-source data acquisition and preprocessing: real-time acquisition of ship power load data, photovoltaic irradiance data, energy storage system SOC value, diesel engine operating parameters and ship navigation state parameters, establishment of unified timestamp time series data set;

[0007] S2, Multimodal data fusion processing: Perform outlier detection and missing data reconstruction on the time series dataset to generate a feature fusion dataset including device health factor labels;

[0008] S3, Dynamic load prediction modeling: Based on the feature fusion dataset, build a hybrid prediction model combining deep learning and time series analysis, output the load prediction curve of the future time window;

[0009] S4, Dynamic power allocation scheme generation: According to the load prediction curve and real-time SOC value, combined with the device aging compensation coefficient, calculate the optimal power allocation ratio of each energy subsystem, and generate a dynamic power allocation scheme;

[0010] S5, Real-time dynamic adjustment and feedback optimization: Based on the deviation value of the real-time operation data of the power system and the power allocation scheme, trigger the dynamic adjustment mechanism to update the power allocation parameters;

[0011] S6, Control instruction generation and execution: Convert the updated power allocation parameters into control instruction sets for each subsystem to achieve coordinated control of the diesel / light / energy storage system.

[0012] Further, the S1 includes:

[0013] S11, Use distributed data acquisition terminals to acquire multi-source data, including ship power load data, photovoltaic irradiance data, energy storage system SOC value, diesel engine operating parameters and ship navigation state parameters;

[0014] S12, Based on the master clock source, build a unified timestamp mechanism, and synchronize the collected multi-source data to generate a time series dataset.

[0015] Further, the S2 includes:

[0016] S21, Based on each type of sensor data sequence in the time series dataset, construct a multimodal observation vector;

[0017] S22, Perform outlier detection on each modal observation vector, and use the sliding statistical threshold method to determine the outlier points;

[0018] S23, For missing values and outliers, use a two-stage reconstruction algorithm to complete data completion;

[0019] S24, Based on the modal observation vector after data completion, introduce a set of device operating state parameters, and use a multi-factor health score model to calculate the device health factor;

[0020] S25, Concatenate the modal data vector after data completion with the corresponding device health factor to generate a feature fusion dataset.

[0021] Further, the sliding statistical threshold method in S22 includes:

[0022] S221, set the length of the sliding window as , and perform statistical calculation on each modal observation vector in the window interval to obtain the local sliding mean and the local sliding standard deviation .

[0023] S222, according to the abnormal threshold coefficient , calculate the decision index , when , the modal observation vector is determined as an abnormal value.

[0024] Further, S3 includes:

[0025] S31, construct the prediction input and output target: based on the feature fusion dataset, set the historical input sequence, and predict the target as the future H-step ship load sequence;

[0026] S32, establish a hybrid prediction model: input the historical input sequence into the hybrid model composed of a bidirectional long short-term memory network and a linear regression layer to predict the future load value.

[0027] Further, S4 includes:

[0028] S41, according to the predicted future load value, estimate the average load power demand and determine the upper limit of the available power of each energy subsystem;

[0029] S42, introduce the device aging compensation coefficient and calculate the distribution ratio: set the aging compensation coefficient of each subsystem as , define the weighted energy supply capacity of each subsystem, and calculate the distribution ratio according to the normalization principle;

[0030] S43, generate a dynamic power distribution scheme: according to the average load prediction value and the weight, determine the distribution power of each energy subsystem;

[0031] construct the final dynamic power distribution scheme output vector.

[0032] Further, S41 includes:

[0033] S411, according to the predicted future load value, calculate the average load power demand;

[0034] S412, set the energy subsystem number as m, and determine the upper limit of the available power of each energy subsystem.

[0035] Further, S5 includes:

[0036] S51, calculate power execution deviation: calculate the theoretical load distribution sum generated at the last moment by the actual total load sampling value of the power system And the total deviation value ;

[0037] S52, introduce dynamic adjustment factor and correct distribution ratio: introduce feedback adjustment factor, correct the current distribution ratio according to the total deviation, and normalize the corrected distribution ratio.

[0038] Further, the S6 comprises:

[0039] S61, construct subsystem target power control vector: based on the normalized corrected distribution ratio, dynamically adjust the optimized power distribution result, and obtain the current running state parameter set of each subsystem ;

[0040] S62, construct control instruction generation function: construct control function for the target power value and state parameter of each subsystem;

[0041] S63, generate and issue control instruction set: combine all control instructions into control instruction set, issue the control instruction set to each subsystem control terminal through the energy management system, and realize synchronous cooperative control.

[0042] The beneficial effects of the present application are:

[0043] The ship diesel / light / energy storage hybrid power system load prediction and dynamic power matching method provided by the present application can fully integrate multi-modal information such as power load data, photovoltaic irradiance, energy storage SOC value, diesel engine operating parameters and sailing state, construct a unified time series data set, realize accurate prediction of future load change trend of the ship, introduce equipment aging compensation mechanism and SOC threshold limit condition, combine deep learning and dynamic feedback optimization algorithm, dynamically generate adaptive power distribution scheme for different energy subsystems, and have stronger real-time and robustness when dealing with sudden load change and equipment state fluctuation.

[0044] The present application further realizes the cooperative operation and fine scheduling among multiple energy subsystems by mapping the optimized power distribution result into the control instruction set of each energy unit, improves the overall response efficiency and energy use efficiency of the system; The control layer supports personalized control instruction generation based on running state, which can effectively avoid problems such as frequent start-stop of diesel engine, excessive charge-discharge of energy storage system, and repeated power disturbance of photovoltaic system, significantly enhances the system operation stability, economy and intelligence level, and is suitable for green energy management demand of new intelligent ships and complex marine working conditions. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description only belong to the present application, and other drawings can also be obtained by those skilled in the art without creative effort.

[0046] Fig. 1 The method flowchart of the embodiment of the present application is shown in the following table.

[0047] Fig. 2 The data fusion diagram of the embodiment of the present application is shown in the following table. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solutions and advantages of the present application more clear, the present application will be further described in detail in combination with specific embodiments.

[0049] As shown in the following table, the ship diesel / light / energy storage hybrid power system load prediction and dynamic power matching method comprises the following steps: Figs. 1-2

[0050] S1, multi-source data acquisition and preprocessing: real-time acquisition of ship power load data, photovoltaic irradiance data, energy storage system SOC value, diesel engine operation parameters and ship navigation state parameters, establishment of a unified timestamp time series data set;

[0051] S2, multi-modal data fusion processing: abnormal value detection and missing data reconstruction are performed on the time series data set to generate a feature fusion data set including device health factor markers;

[0052] S3, dynamic load prediction modeling: based on the feature fusion data set, a hybrid prediction model combining deep learning and time series analysis is constructed to output a load prediction curve for a future time window;

[0053] S4, dynamic power distribution scheme generation: based on the load prediction curve and real-time SOC value, combined with the device aging compensation coefficient, the optimal power distribution ratio of each energy subsystem is calculated to generate a dynamic power distribution scheme;

[0054] S5, real-time dynamic adjustment and feedback optimization: based on the deviation value of the real-time operation data of the power system and the power distribution scheme, a dynamic adjustment mechanism is triggered to update the power distribution parameters;

[0055] S6, control instruction generation and execution: the updated power distribution parameters are converted into a control instruction set for each subsystem to realize the collaborative control of the diesel / light / energy storage system.

[0056] S1 comprises:

[0057] ​S11, acquire multi-source data by using a distributed data acquisition terminal, including ship power load data, photovoltaic irradiance data, energy storage system SOC value, diesel engine operating parameter and ship navigation state parameter, specifically including:

[0058] (1) acquire voltage, current and active power of the ship's main power distribution network through the AC sampling module to form the power load data ;

[0059] (2) acquire the light intensity on the surface of the photovoltaic panel through the light sensor , and calculate the equivalent available photovoltaic power by combining the conversion efficiency of the photovoltaic module , which is expressed as:

[0060] wherein A is the total area of the photovoltaic module;

[0061] (3) acquire the state value by acquiring the ratio of the remaining power of the energy storage system to the rated capacity through the energy storage management system ;

[0062] (4) acquire the diesel engine operating parameter vector by acquiring the speed , fuel injection amount and fuel consumption rate of the diesel engine through the diesel engine ECU module ;

[0063] (5) acquire the navigation state vector by acquiring the speed , heading angle and current sea state level through the navigation positioning and inertial navigation system ; ;

[0064] S12, based on a high-precision master clock source, a unified timestamp mechanism is constructed to synchronize the collected multi-source data, and a time series data set is generated, which is expressed as:

[0065] ;

[0066] wherein t is the unified timestamp, the interval is a fixed time step , the high-precision master clock source sets the ship power control center as the master clock node of the whole system, and uses the synchronization network clock protocol (PTP, Precision Time Protocol) to synchronize the time of all distributed data acquisition terminals.

[0067] S2 includes:

[0068] S21, based on various types of sensor data sequences in the time series data set, a multi-modal observation vector is constructed, which is expressed as:

[0069] ;

[0070] in, This represents the multimodal observation vector acquired at time t, where d is the modal dimension. This represents the observation value of the i-th mode;

[0071] S22, perform outlier detection for each modal observation vector, using a sliding statistical threshold method to determine outliers. Outliers are marked as outliers when certain conditions are met, as follows:

[0072] ;

[0073] in, For mode i in the sliding window The mean within, For the corresponding standard deviation, The threshold value for the anomaly coefficient is empirically set between 2 and 3. The length of the sliding window. Let i be the observation vector of the i-th mode;

[0074] S23. For missing and outlier values, a two-stage reconstruction algorithm is used to complete the data. The specific steps include:

[0075] (1) The first stage uses linear interpolation, expressed as:

[0076] ;

[0077] in, and , These are adjacent valid observation points;

[0078] (2) In the second stage, the interpolation results are corrected using a first-order autoregressive model, as shown below:

[0079] ;

[0080] in, It is an autoregressive smoothing factor;

[0081] S24, Modal observation vectors based on data completion Introduce equipment operating status parameter set , The health factors of the equipment are calculated using a multi-factor health scoring model, and are expressed as follows:

[0082] ;

[0083] in, This is the device's Health Index; the lower the value, the lower the health level. is the jth equipment running state parameter, is the jth equipment running state parameter health factor weighting coefficient, satisfying is the standard deviation normalization processing function, , is the historical mean of the state parameter , is the corresponding standard deviation, and the normalized result obeys a normal distribution with a mean of 0 and a variance of 1;

[0084] S25, the modal data vector after data completion is spliced with the corresponding equipment health factor to generate a feature fusion data set, denoted as:

[0085] .

[0086] The sliding statistical threshold method adopted in S22 includes:

[0087] S221, set the sliding window length to , and perform statistical calculation on each modal observation vector in the window interval to obtain the local sliding mean and the local sliding standard deviation , denoted as:

[0088] ;

[0089] ;

[0090] wherein, is the observation value collected by the ith modal observation vector at time t, is the sliding mean of the ith modal observation vector in the window , is the sliding standard deviation of the ith modal observation vector in the window , and k is the time index variable in the window, taking a value range of , is the sliding window length;

[0091] S222, according to the abnormal threshold coefficient , calculate the decision index , when , the modal observation vector is determined as an abnormal value, denoted as:

[0092] .

[0093] S3 includes:

[0094] S31, constructing prediction input and output target: based on the feature fusion dataset, setting the historical input sequence, and predicting the target as the future H-step ship load sequence, denoted as:

[0095] ;

[0096] wherein, is the feature fusion vector at the t-th time point, including the reconstructed multi-modal observation value and the equipment health factor, is the historical feature sequence, L represents the length of the backtracking time window, i.e., the number of historical steps for modeling, and d is the dimension of the fused features;

[0097] ;

[0098] wherein, represents the predicted value of the load at the t+h time point, is the future H-step ship load sequence;

[0099] S32, establishing a hybrid prediction model: inputting the historical input sequence into a hybrid model composed of a bidirectional long short-term memory network (Bi-LSTM) and a linear regression layer to predict the future load value, the specific steps including:

[0100] (1) the Bi-LSTM layer extracts the time series dynamic features, denoted as:

[0101] ;

[0102] wherein, is the hidden state vector at the t-th time point, encoding the evolution trend of the historical features, is the dimension of the hidden state;

[0103] (2) the linear mapping layer predicts the future load value, denoted as:

[0104] ;

[0105] wherein, is the weight vector of the h-th step prediction, is the corresponding bias term.

[0106] S4 includes:

[0107] S41, according to the predicted future load value, estimating the average load power demand and determining the upper limit of the available power of each energy subsystem;

[0108] S42, introducing the equipment aging compensation coefficient and calculating the distribution ratio: assuming that the aging compensation coefficient of each subsystem is , the weighted energy supply capacity of each subsystem is defined, denoted as:

[0109] ;

[0110] According to the normalization principle, the distribution ratio is calculated , which is expressed as:

[0111] ;

[0112] wherein m is an index representing the energy subsystem number, taking values of 1 (diesel engine system), 2 (photovoltaic system), and 3 (energy storage system), and n is an auxiliary subsystem number in the denominator summation, is the weighted available power of the mth subsystem, is the maximum available power of the mth subsystem at the current time t, determined by the system state or resource conditions, wherein the SOC limit needs to be considered for the energy storage system, is the power distribution ratio of the mth subsystem;

[0113] S43, generating a dynamic power distribution scheme: according to the average load prediction value and the weight, the distribution power of each energy subsystem is determined, which is expressed as:

[0114] ;

[0115] The final dynamic power distribution scheme output vector is constructed, which is expressed as:

[0116] ;

[0117] wherein is the actual energy supply target value of the mth subsystem in this round of distribution, is the average load power demand in the prediction time window, output by the prediction module, is the final power distribution scheme output result of the three subsystems.

[0118] S41 includes:

[0119] S411, according to the predicted future load value , the average load power demand is calculated, which is expressed as:

[0120] ;

[0121] S412, assuming that the energy subsystem number is m (m=1, 2, 3), the upper limit of the available power of each energy subsystem is determined, which is expressed as:

[0122] ;

[0123] wherein is the current state of charge of the energy storage system, is a SOC state control factor, , is a set SOC lower threshold value, .

[0124] S5 includes:

[0125] S51, calculate power execution deviation: calculate the actual total load sampling value of the power system , calculate the theoretical load distribution sum generated at the last time and the total deviation value , expressed as:

[0126] ;

[0127] ;

[0128] Wherein, if , it indicates that the current power distribution is insufficient, and the output needs to be increased, if , it indicates that the distribution is excessive, and the output needs to be appropriately reduced; is the actual total load sampling value of the power system at the current time t, which is calculated based on the ship power load data collected in S1, , is the real-time active power value of the lth load node, and N is the number of load nodes currently participating in collection;

[0129] S52, introduce dynamic adjustment factor and correct distribution ratio: introduce feedback adjustment factor, correct the current distribution ratio according to the total deviation, and normalize the corrected distribution ratio, expressed as:

[0130] ;

[0131] Wherein, is the weighted power supply capacity of the subsystem m at the current time, indicates that the deviation is distributed to each subsystem after being weighted according to the power supply capacity, is the normalization reference, is a dynamic adjustment coefficient, , used to control the feedback response rate, is the original distribution ratio of the subsystem m at the current time, is the unnormalized distribution ratio after one adjustment;

[0132] ;

[0133] Wherein, is the normalized final distribution ratio, is the variable index.

[0134] S6 includes:

[0135] S61, constructing a subsystem target power control vector: based on the normalized correction allocation ratio, dynamically adjusting the optimized power allocation result, and obtaining the current running state parameter set of each subsystem , expressed as:

[0136] ;

[0137] ;

[0138] wherein, is the target power value of the mth energy subsystem allocated at the current time t, is the energy subsystem number, corresponding to the diesel engine system, photovoltaic system and energy storage system respectively;

[0139] ;

[0140] wherein, is the mth state parameter, including diesel engine speed, energy storage SOC and voltage, is the number of state parameters of the mth subsystem;

[0141] S62, constructing a control instruction generation function: for the target power value and state parameter of each subsystem, constructing a control function, expressed as:

[0142] ;

[0143] wherein, is the control instruction amount of the mth subsystem, including speed set value, MPPT reference point and converter power target, is a subsystem-specific control mapping function;

[0144] S63, generating and issuing control instruction set: combining all control instructions into a control instruction set, through the energy management system, issuing the control instruction set to each subsystem control terminal respectively to realize synchronous and collaborative control, expressed as:

[0145] .

[0146] ​​Those skilled in the art should understand that the above discussion of any embodiment is only intended to be exemplary in nature and is not intended to suggest that the present application is limited to these examples; under the concept of the present application, the above embodiments or technical features among different embodiments can be combined, steps can be implemented in any order, and there are many other variations of different aspects of the present application as described above, which are not provided in details for the sake of brevity.

Claims

1. A method for load forecasting and dynamic power matching of a shipboard diesel / solar / storage hybrid power system, characterized in that, Includes the following steps: S1, Multi-source data acquisition and preprocessing: Real-time acquisition of ship power load data, photovoltaic irradiance data, energy storage system SOC value, diesel engine operating parameters and ship navigation status parameters, to establish a time series dataset with a unified timestamp; S2, Multimodal data fusion processing: outlier detection and missing data reconstruction are performed on the time series dataset to generate a feature fusion dataset including equipment health factor labels; S3, Dynamic load forecasting modeling: Based on the aforementioned feature fusion dataset, a hybrid forecasting model combining deep learning and time series analysis is constructed to output load forecast curves for future time windows; S4, Dynamic Power Allocation Scheme Generation: Based on the load forecast curve and real-time SOC value, combined with the equipment aging compensation coefficient, the optimal power allocation ratio for each energy subsystem is calculated, and a dynamic power allocation scheme is generated; specifically including: S41, based on the predicted future load values, estimate the average load power demand and determine the upper limit of available power for each energy subsystem; S42, Introduce equipment aging compensation coefficients and calculate allocation ratios: Let the aging compensation coefficient for each subsystem be... Define the weighted energy supply capacity of each subsystem and calculate the allocation ratio according to the normalization principle; S43, Generate dynamic power allocation scheme: Based on the average load forecast value and weight, determine the allocated power of each energy subsystem and construct the final dynamic power allocation scheme output vector; S41 includes: S411, Calculate the average load power demand based on the predicted future load values; S412, let the energy subsystem be numbered as Determine the upper limit of available power for each energy subsystem; S5, Real-time dynamic adjustment and feedback optimization: Based on the deviation between the real-time operating data of the power system and the power allocation scheme, a dynamic adjustment mechanism is triggered to update the power allocation parameters; S6, Control Command Generation and Execution: Converts the updated power allocation parameters into control command sets for each subsystem to achieve coordinated control of the diesel / photovoltaic / storage system.

2. The method for load forecasting and dynamic power matching of a shipboard diesel / solar / storage hybrid power system according to claim 1, characterized in that, S1 includes: S11 uses distributed data acquisition terminals to acquire multi-source data, including ship power load data, photovoltaic irradiance data, energy storage system SOC value, diesel engine operating parameters and ship navigation status parameters; S12, based on the master clock source, constructs a unified timestamp mechanism to synchronize the collected multi-source data and generate a time series dataset.

3. The method for load forecasting and dynamic power matching of a shipboard diesel / solar / storage hybrid power system according to claim 2, characterized in that, S2 includes: S21, Construct multimodal observation vectors based on various sensor data sequences in the time series dataset; S22, perform outlier detection for each modal observation vector and use the sliding statistical threshold method to determine outliers; S23. For missing and outlier values, a two-stage reconstruction algorithm is used to complete the data. S24. Based on the modal observation vector after data completion, the equipment operating status parameter set is introduced, and a multi-factor health scoring model is used to calculate the equipment health factors. S25, the modal data vector after data completion is concatenated with the corresponding device health factors to generate a feature fusion dataset.

4. The method for load forecasting and dynamic power matching of a ship diesel / solar / storage hybrid power system according to claim 3, characterized in that, The sliding statistical threshold method used in S22 includes: S221, Set the sliding window length to... For each modal observation vector within the window interval Perform statistical calculations within the region to obtain the local moving average. and local sliding standard deviation ; S222, based on the abnormal threshold coefficient Calculate the judgment index ,when When this happens, the modal observation vector is identified as an outlier.

5. The method for load forecasting and dynamic power matching of a shipboard diesel / solar / storage hybrid power system according to claim 4, characterized in that, S3 includes: S31, Constructing the Predicted Input and Output Objective: Based on the feature fusion dataset, define the historical input sequence and predict the target as the future. Step-by-step ship load sequence; S32, Establish a hybrid prediction model: Input the historical input sequence into a hybrid model consisting of a bidirectional long short-term memory network and a linear regression layer to predict future load values.

6. The method for load forecasting and dynamic power matching of a shipboard diesel / solar / storage hybrid power system according to claim 1, characterized in that, S5 includes: S51, Calculate the power execution deviation: Calculate the sum of the theoretical load allocation generated at the previous moment using the actual total load sample value of the power system. and total deviation value ; S52, Introduce a dynamic adjustment factor and correct the allocation ratio: Introduce a feedback adjustment factor to correct the current allocation ratio based on the total deviation, and normalize the corrected allocation ratio.

7. The method for load forecasting and dynamic power matching of a shipboard diesel / solar / storage hybrid power system according to claim 6, characterized in that, S6 includes: S61, Construct the target power control vector for the subsystem: Based on the normalized corrected allocation ratio, dynamically adjust the optimized power allocation result and obtain the current operating state parameter set of each subsystem. ; S62, Construct control command generation function: Construct control function for each subsystem's target power value and state parameters; S63, Generate and issue control instruction set: Combine all control instructions into a control instruction set, and issue the control instruction set to the control terminals of each subsystem through the energy management system to achieve synchronous and coordinated control.

Images