Intelligent scheduling method and system for heterogeneous load energy supply and demand in building

Abstract

The invention discloses an intelligent scheduling method and system for heterogeneous load energy supply and demand in a building, and the method comprises the steps: building a heterogeneous load dynamic recognition model, taking heterogeneous load data after feature extraction as a training sample set to train the model, obtaining a recognition model used for recognizing the type of equipment, and carrying out the recognition of the type of the equipment; and determining the type of the to-be-detected device based on the identification model, thereby determining a communication protocol type based on the device type, establishing a communication connection, predicting an energy demand by using a time sequence algorithm based on identification and communication data, and making an energy management strategy in combination with heterogeneous load data. According to the method, accurate identification of heterogeneous loads is realized through multi-dimensional data acquisition and feature extraction. And communication protocols adopted by various heterogeneous loads can be automatically detected and identified according to the equipment types, rapid adaptation is completed, and an intelligent energy management strategy is further rapidly formulated.

Description

Technical Field

[0001] The present invention belongs to the field of energy management and intelligent control of zero-carbon park buildings, and particularly relates to a method and system for intelligent scheduling of energy supply and demand of heterogeneous loads in buildings, aiming to effectively manage various types of electrical equipment in public buildings and improve energy utilization efficiency and intelligent operation level. Background Art

[0002] With the increasing diversification and intelligent development of modern zero-carbon park building functions, a large number of electrical equipment of different types, brands and manufacturing eras are connected to the building, namely heterogeneous loads. These heterogeneous loads include but are not limited to various air conditioners (such as central air conditioners, split air conditioners, multi-split air conditioners and other different systems and different brands of products), water boilers (different heating principles and capacity specifications), charging piles (AC, DC, different power levels and communication protocols), etc.

[0003] In the traditional zero-carbon campus building energy management system, there are many problems due to the lack of accurate identification and effective integration of these heterogeneous loads. On the one hand, it is impossible to accurately grasp the detailed information of various heterogeneous loads in real time, such as type, power demand changes, energy consumption characteristics, etc., which makes it difficult to make refined predictions and optimize the overall energy consumption of the building. For example, during peak hours of electricity consumption, due to the lack of clarity on the actual power demand and operating status of each load, unreasonable power distribution may occur, some key equipment is underpowered, and some non-essential equipment consumes too much energy, affecting the normal operation of building functions and causing energy waste. On the other hand, heterogeneous loads usually use different communication protocols to interact with the control system, which makes it a huge challenge to build a unified building energy management platform. The traditional fixed communication protocol configuration method is difficult to adapt to the access of many devices with different protocols, resulting in poor device access compatibility, incomplete or inaccurate data collection, and thus unable to achieve centralized monitoring and intelligent coordinated control of heterogeneous loads. For example, when a new brand or model of equipment needs to be connected to the existing energy management system, it often takes a lot of manpower and time to reconfigure and debug the system, which seriously restricts the scalability and intelligent upgrade of the zero-carbon campus building energy management system. Summary of the invention

[0004] In view of the above problems, the technical solution adopted by the present invention is: a method for intelligent scheduling of energy supply and demand of heterogeneous loads in a building, the method comprising the following steps:

[0005] Acquire different types of heterogeneous load data, wherein the heterogeneous load data includes the equipment type of the load and the corresponding electrical parameters and operation data;

[0006] Establishing a dynamic identification model for heterogeneous loads, using the heterogeneous load data after feature extraction as a training sample set to train the model, and obtaining an identification model for identifying equipment types;

[0007] Obtaining heterogeneous load data of the equipment to be detected connected to the building, and determining the type of the equipment to be detected based on the recognition model after feature extraction;

[0008] Determine the communication protocol type based on the device type and establish a communication connection;

[0009] Based on identification and communication data, time series algorithms are used to predict energy demand, and energy management strategies are formulated in combination with heterogeneous load data.

[0010] Optionally, the feature extraction comprises the following steps:

[0011] Calculating active power, reactive power, power factor and current waveform characteristics based on the electrical parameters;

[0012] A feature vector is constructed for the operation data according to the equipment type of the heterogeneous load.

[0013] Optionally, the step of establishing a dynamic identification model for heterogeneous loads includes:

[0014] The support vector machine (SVM) algorithm is used to build a dynamic identification model for heterogeneous loads.

[0015] The collected heterogeneous load data is used as the training sample set in is the feature vector of the i-th sample, y i is the corresponding device type label;

[0016] The radial basis function is selected as the kernel function to train the SVM model, and the decision function parameters are solved by the sequential minimum optimization algorithm.

[0017] Optionally, the step of formulating an energy management strategy includes:

[0018] Conduct statistical analysis based on heterogeneous load data from different types of equipment;

[0019] Determine the energy consumption characteristics of different types of equipment;

[0020] After determining the type of equipment according to the identification model, synchronously outputting the power demand value and energy consumption characteristics of the equipment;

[0021] Based on historical energy consumption data, real-time power demand values ​​of heterogeneous loads, and external environmental factors, a multivariate time series prediction model is constructed, and a deep learning algorithm is used to predict the energy demand of each heterogeneous load.

[0022] According to the energy forecast structure and the importance, energy consumption characteristics and operating status of various heterogeneous loads, a dynamic priority allocation algorithm is used to formulate an energy management strategy.

[0023] Optionally, the method further comprises the following steps:

[0024] Regularly testing and evaluating the recognition model;

[0025] If the accuracy of the recognition model is lower than the set value, analyze and determine the cause;

[0026] Obtain new heterogeneous load data based on the analyzed cause type, extract features and enter them into the training set;

[0027] Keep the algorithm unchanged and adjust the parameters to retrain the model based on the updated training set.

[0028] Optionally, the step of determining the type of communication protocol based on the device type and establishing a communication connection includes:

[0029] Design different communication detection instructions for the device types of heterogeneous loads, wherein the communication detection instructions include operation data request instructions corresponding to the device types, and the instruction formats include multiple communication protocol formats and cover different communication parameters;

[0030] Detect physical connections and communication signals when devices are connected, and determine identifiability and communication capabilities;

[0031] If the judgment is passed, a corresponding communication detection instruction is sent to it based on the device type.

[0032] Optionally, the step of determining the type of communication protocol based on the characteristic information of the response data includes:

[0033] Conducting in-depth analysis on the collected response data to extract protocol feature information;

[0034] Comparing the extracted protocol feature information with a pre-stored communication protocol feature library, and using a similarity algorithm to calculate the similarity between the protocol feature information and each known protocol feature;

[0035] Determine the communication protocol type based on the similarity results.

[0036] Optionally, the step of calculating the similarity includes:

[0037] Determine the weight values ​​of each feature in different protocols;

[0038] Comparing the protocol feature information extracted from the response data with features of corresponding multiple protocols;

[0039] The similarity is calculated based on the weights of the same features.

[0040] And, an intelligent dispatching system for energy supply and demand of heterogeneous loads in a building, the system comprising:

[0041] A data acquisition module is used to obtain heterogeneous load data, wherein the heterogeneous load data includes the equipment type of the load and the corresponding electrical parameters and operation data;

[0042] A feature extraction module, used for extracting features from the heterogeneous load data;

[0043] A dynamic identification model module is used to establish a dynamic identification model for heterogeneous loads, and train the model using the heterogeneous load data after feature extraction as a training sample set to obtain an identification model for identifying the type of equipment; and determine the type of the equipment to be detected based on the identification model and the heterogeneous load data of the equipment to be retrieved;

[0044] An adaptive communication configuration module, used to determine the type of communication protocol according to the device type and establish a communication connection;

[0045] The energy management and control module is used to predict energy demand based on identification and communication data using time series algorithms, and to formulate energy management strategies in combination with heterogeneous load data.

[0046] Optionally, the dynamic identification model is also used to determine the power demand value and energy consumption characteristics of the equipment based on the equipment information; the energy management and control module is specifically used to construct a multivariate time series prediction model based on historical energy consumption data, real-time power demand values ​​of heterogeneous loads and external environmental factors, and use a deep learning algorithm to predict the energy demand of each heterogeneous load; it is also used to formulate an energy management strategy based on the energy prediction structure and the importance, energy consumption characteristics and operating status of each heterogeneous load using a dynamic priority allocation algorithm.

[0047] The present invention adopts the above technical solution, so that it has the following beneficial effects: through multi-dimensional data collection and feature extraction, accurate identification of heterogeneous loads is achieved. It can automatically detect and identify the communication protocols used by various heterogeneous loads according to the device type, complete rapid adaptation, and further obtain real-time load demand and energy supply conditions, and formulate intelligent energy management strategies.

[0048] Through multi-dimensional data collection and feature extraction, accurate identification of heterogeneous loads is achieved. Compared with the traditional identification method that only relies on a single electrical parameter or simple rule judgment, the accuracy is significantly improved. This enables the building energy management system to accurately know the type and operating status of each load, providing a solid foundation for subsequent refined management. At the same time, by quickly and accurately establishing communication connections through the identified heterogeneous load types, the compatibility and scalability of the system are greatly improved, and the access difficulties caused by the mismatch of communication protocols when facing new equipment or niche equipment in traditional systems are effectively solved, reducing the cost and difficulty of system integration and maintenance.

[0049] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 A flow chart of a method for dynamically identifying heterogeneous loads according to an embodiment of the present invention is shown;

[0052] Figure 2 A flow chart of an adaptive configuration method according to an embodiment of the present invention is shown;

[0053] Figure 3 A block diagram of a system for dynamic identification and adaptive configuration of heterogeneous loads according to an embodiment of the present invention is shown;

[0054] Figure 4 A block diagram of a heterogeneous load energy supply and demand intelligent scheduling system according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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.

[0056] Embodiment 1

[0057] like Figure 1 As shown, the method for dynamically identifying heterogeneous loads in a building according to an embodiment of the present invention comprises the following steps:

[0058] Step S100 Data Collection: Data collection nodes are deployed in each power consumption area of ​​the zero-carbon park building. These nodes are equipped with high-precision current transformers, voltage transformers, and collection modules for different equipment communication interfaces. The current I of each phase is collected at a specific sampling frequency (such as 1000Hz). a (t), I b (t), I c (t), voltage U a (t), U b (t), U c (t) and the corresponding timestamp t. At the same time, the internal operating status data of the device is collected through the communication interface, such as the temperature setting value T of the smart air conditioner set (t), operating mode M(t) (where the cooling mode can be set to M=1, the heating mode to M=2, etc.), the charging current I charge (t), charging voltage U charge (t), remaining battery capacity C charge (t) and other information.

[0059] Step S101: Data preprocessing: First, data cleaning is performed, and the current threshold range (such as I min =-50A to I max =50A) and voltage threshold range (such as U min =190V to U max =250V), and remove abnormal data that exceeds the threshold, which may be caused by sensor failure or electromagnetic interference. Then, the data is synchronized and accurately timestamped to ensure that data from different sources (electrical parameter data and internal operating status data) accurately correspond on the time axis, so that the subsequent feature extraction and analysis steps can accurately process various data features at the same time.

[0060] Step S102: Feature extraction: extract multi-dimensional features from the collected and pre-processed data. In terms of electrical feature extraction, the three-phase active power P(t) is calculated using the formula: The time window T is set to 1s. They are the three-phase power factors, which are obtained by measuring the phase difference between voltage and current. For single-phase equipment, the active power is simplified to Calculate three-phase reactive power Apparent power (three-phase) or S(t)=U(t)I(t)(single-phase), U line (t) is the line voltage, Iline (t) is the line current, and the power factor For the current waveform characteristics, the current signal I(t) is subjected to fast Fourier transform (FFT) to obtain the spectrum I(f) and calculate the harmonic distortion rate (THD) Among them I h is the hth harmonic current amplitude, I1 is the fundamental current amplitude, and the kurtosis of the current waveform is calculated at the same time and skewness is the number of sampling points in the time window, is the current mean, σ I (t) is the current standard deviation.

[0061] In terms of operating data feature extraction, a vector containing key state information is constructed based on the device type. For example, a feature vector is constructed for air conditioners. Where ΔT(t) = T room (t)-T set (t) is the temperature difference between indoor and outdoor; construct feature vector for charging pile P charge (t) = I charge (t)U charge (t) is the charging power.

[0062] Step S103: Model training and identification: Use the support vector machine (SVM) algorithm to build a dynamic identification model for heterogeneous loads. Collect a large amount of heterogeneous load data covering different types and characteristics (including the electrical characteristics and operation characteristics data extracted above) as a training sample set. in is the feature vector of the i-th sample, y i is the corresponding device type label (e.g. air conditioner has a specific label value y=1, charging pile has another label value y=2, etc.). Use radial basis function (RBF) The SVM model is trained as the kernel function, and the decision function parameters are solved by the sequential minimum optimization algorithm (SMO). In this process, the kernel function parameter γ is set to 0.08 and the penalty factor C is set to 8. These parameters are determined through multiple experiments and verifications, which can enable the model to have good generalization ability and recognition accuracy in the current scenario.

[0063] During the model operation, the feature vector extracted in real time is input into the trained SVM model, and the model outputs the corresponding heterogeneous load type label ypre d And related power demand estimation value and energy consumption characteristic parameters. The power demand estimation value is obtained by establishing a mapping relationship based on the power calculated by the heterogeneous load equipment data in the above training recognition model; the energy consumption characteristic parameters include the energy efficiency ratio range, which is obtained by statistical analysis of the historical data of heterogeneous loads of different types of equipment. For example, if the model outputs yp, ed The corresponding air-conditioning equipment type label further provides information such as the current power demand estimate of the air-conditioning equipment and the energy efficiency ratio range under the current operating mode, providing accurate data support for subsequent energy management and adaptive configuration.

[0064] Step S104 Model evaluation and optimization: Regularly use independent test data sets to evaluate the model's recognition accuracy, recall rate, F1 value and other performance indicators. The test data set should contain heterogeneous load data of various types and under different operating conditions, and should be independent of the training data set. If the model performance declines (such as accuracy below 92%), the cause will be analyzed in depth, which may be due to the emergence of new types of equipment, changes in the building's power environment and other factors. For such situations, collect more data on new types of equipment and add them to the training set, retrain the model, and keep the SVM algorithm unchanged. Only adjust the values ​​of the kernel function parameter γ and the penalty factor C, and use cross-validation again to determine the best parameter combination to continuously improve the model's dynamic recognition ability for heterogeneous loads.

[0065] In particular, the incremental learning mechanism step S105 is introduced: when new data flows in, its feature vector is added to the training sample set. The distance between the new data point and the existing support vector is calculated. If the distance is less than the threshold (set as d th =0.5), it is considered to have a greater impact on the model. Only for these relevant support vectors and new data points, the incremental learning algorithm is used to update the model parameters, reducing the consumption of computing resources and quickly adapting to load changes.

[0066] Embodiment 2:

[0067] See also Figure 2 As shown, an embodiment of the present invention provides an efficient and adaptive communication protocol configuration method. After accurately identifying heterogeneous loads through the accurate and comprehensive heterogeneous load dynamic identification method provided in the above embodiment 1, the communication protocol configuration process is started according to the identified load type information, which is as follows:

[0068] Step S200: Device access detection: When heterogeneous loads of zero-carbon park buildings such as central air-conditioning systems, multi-brand electric water heaters, and charging piles of different specifications try to access the zero-carbon park building energy management system, the system first starts the access detection procedure. For physical connection inspection, for central air-conditioning, check the connection stability of its power supply line and control line, and use resistance detection equipment to ensure that the line resistance is within the normal range (for example, 0.1-10 ohms), without short circuit (resistance approaches 0 ohms) or open circuit (resistance approaches infinity); for electric water heaters, check the connection between its heating element and the power supply, as well as the connection line between the temperature control device and the communication module; for charging piles, check the integrity of the connection line between the charging gun and the charging device body and the connectivity with the external communication line. At the same time, the signal detection device is used to detect whether each device has a communication signal. For example, if the controller of the central air-conditioning works normally, it should send a communication signal pulse of a specific frequency (such as 10kHz) every 5 seconds. If the signal is not detected, it is preliminarily judged that the device communication may be abnormal or not in a recognizable state.

[0069] Step S201 protocol detection sending: the system sends specially designed standard communication detection instructions for different types of heterogeneous loads. For central air conditioners, instructions including temperature setting query, operation mode acquisition, fan speed request, etc. are sent. The instruction format covers a variety of common central air conditioner communication protocol formats, such as 03 function code instructions under Modbus RTU protocol format (used to read multiple register data, such as temperature setting value register), 06 function code instructions (used for single register writing, which can be used to test whether the control instruction can be received by the device), and each function code instruction is sent at different baud rates (such as 9600bps, 19200bps, 38400bps) and data bits (such as 8 data bits), stop bits (such as 1 stop bit) Combination sent; for electric water heaters, including water temperature setting query, heating power acquisition, etc. Instructions, including simple text protocol format (such as "GET TEMP The "SETTING" command is used to obtain the temperature setting) and similar protocol formats customized by some manufacturers, and the baud rate is sent between 2400bps-115200bps; for charging piles, the charging status query, charging power limit setting and other commands are sent, and the specific message frame format in the national standard charging pile protocol (GB / T 27930-2015) and the commands compatible with some European and American standard protocol formats are used to detect at different baud rates (such as 57600bps, 115200bps).

[0070] Step S202 Data sniffing and capture: while sending the detection instruction, start the high-sensitivity data sniffing function. The sampling frequency of the data capture module is set to 1MHz to ensure that weak communication signals and data frames of various complex formats can be captured. For example, when sniffing the communication interface of the central air conditioner, after sending the 03 function code instruction, the returned data frame is captured. The data frame may contain the start bit "55", the stop bit "AA", and the check bit is two bytes of data calculated by the CRC-16 algorithm. The data length depends on the number of registers queried. If 5 registers are queried, the data length is 11 bytes (including function code, register address, number of data bytes, etc.). For electric water heaters, if a simple text protocol format is used, the captured data may be a data string that starts with a specific character (such as ") and ends with a carriage return and line feed character ("\r\n"), such as TEMP_SETTING: 50\r\n", indicating that the temperature is set to 50 degrees Celsius. For charging piles, the charging status data frame captured according to the national standard protocol contains a specific status code field (such as 0x01 for charging, 0x02 for charging completion) and data fields such as charging current and voltage. The data frame length is fixed at 25 bytes.

[0071] Step S203 protocol feature extraction: conduct in-depth analysis of the captured response data and extract key protocol feature information. Such as frame format (start, stop, check bit), length, content features (field meaning distribution), build protocol framework, etc. Taking central air conditioning as an example, if the data frame format is Modbus RTU protocol, determine the start bit as "55", the stop bit as "AA", the check bit as the result of CRC-16 algorithm calculation, and the data length is determined according to the instruction response situation; in terms of data content features, analyze the operation meaning corresponding to the function code, such as function code 03 corresponds to reading register data, and determine whether the temperature setting value or operation mode is read according to the register address. For example, register address 0x0001 corresponds to the temperature setting value register, and the temperature setting value of the central air conditioner can be obtained by parsing the register data. For electric water heaters, determine the specific command format and data separator of the text protocol, such as "$" at the beginning and "\r\n" at the end, and ":" as the data separator, and determine the water temperature setting and other information by parsing the command string. For charging piles, extract the position and length information of data fields such as status code, charging current, voltage, etc. in the national standard protocol data frame. For example, the status code is located in the 3rd byte of the data frame, the charging current data is located in the 5th to 8th bytes (expressed as a 4-byte floating point number), and the voltage data is located in the 9th to 12th bytes.

[0072] Step S204 protocol feature comparison: compare and match the extracted communication protocol feature information with a variety of common communication protocol feature libraries pre-stored in the system. A weighted similarity algorithm is used to calculate the similarity between the captured protocol features and each known protocol feature. For the Modbus RTU protocol feature comparison of central air conditioners, the weights of key features such as data frame format and verification method are set to 0.6, and the weights of auxiliary features such as function code meaning are set to 0.4. For example, if the captured data frame format completely matches the Modbus RTU protocol, the function code meaning is also correctly parsed, and the check digit is calculated correctly, the similarity can reach more than 0.9; if the data frame format partially matches and the function code meaning is partially correct, the similarity is between 0.5-0.7. When the similarity exceeds the set threshold (such as 0.7), the type of communication protocol adopted by the device is determined. A similar weighted comparison method is also used for electric water heaters and charging piles. For example, the specific command format weight of the electric water heater text protocol is 0.7, and the data separator weight is 0.3; the status code position and meaning weight of the charging pile Chinese standard protocol is 0.5, the charging current and voltage data field format weight is 0.3, and the weight of other auxiliary features is 0.2.

[0073] Step S205: Manual intervention: If the type of communication protocol cannot be determined after comparison and matching, such as a new type of energy-saving air conditioner adopts a new encryption communication protocol, manual intervention is prompted. Professional communication technicians use professional communication analysis tools, such as high-precision protocol analyzers (such as Agilent N4010A protocol analyzer) and oscilloscopes (such as Tektronix TDS2000 series oscilloscopes), to further analyze the equipment communication protocol. The technicians capture the communication data of the air conditioner through the protocol analyzer, analyze its encryption algorithm and data transmission rules, and enter the analysis results into the system in detail. The system automatically updates the protocol feature library so that it can identify this type of protocol in the future, and continuously expand the system's recognition ability for unknown protocols.

[0074] Step S206 Adaptive configuration generation: According to the determined communication protocol type, the system automatically generates corresponding communication configuration parameters. For central air conditioners, if it is determined to be Modbus RTU protocol, the communication port is set to RS485 port, the baud rate is set to 19200bps according to the device response (if the device responds fastest and stably at this baud rate), and the data format is set to 8 data bits, no check bit (if the device adopts this data format), and 1 stop bit; for electric water heaters, if it is a simple text protocol, the communication port selects a suitable serial port (such as COM3) according to the device connection, the baud rate is set to 9600bps, the data format is text format, and no special check bit is set; for charging piles, if it is a national standard protocol, the communication port is set to an Ethernet port (if supported) or an RS485 port, the baud rate is set to 115200bps, the data format is set according to the message frame format specified in the national standard protocol, and the communication timeout is set to 500 milliseconds (determined according to the device response characteristics) to avoid long system waiting due to device failure or communication delay.

[0075] Step S207: Establishing a communication connection: Apply the generated communication configuration parameters to the process of establishing a communication connection with heterogeneous loads. The system attempts to establish a stable communication connection with the device according to the set parameters. During this process, it continuously monitors the connection status, such as whether the connection is successfully established and whether the data transmission is normal. If the connection establishment fails, the system automatically performs troubleshooting to check whether the parameter settings are correct and whether the device responds normally. For example, if the connection with the central air conditioner fails, first check whether the connection line of the RS485 port is loose, and then check whether the baud rate, data format and other parameters match the actual requirements of the device. By sending specific test instructions to the device (such as 06 function code instructions to set a fixed temperature value) and checking whether the device has corresponding actions or response data returned, make corresponding adjustments based on the troubleshooting results, such as resetting the baud rate or checking the line and then try to establish the connection again.

[0076] Step S208 Communication status monitoring: After the communication connection is successfully established, a communication status monitoring mechanism is established. This mechanism monitors the quality of the communication connection and the accuracy of data transmission in real time, and determines whether there are abnormal conditions such as data packet loss and communication interruption by sending a heartbeat packet every 3 seconds (such as sending a specific empty command for the central air conditioner, which is only used to detect whether the connection is alive) and checking the data verification results. If a communication abnormality is found, the system automatically adjusts the communication parameters. For example, in the communication monitoring of the charging pile, if the data packet loss rate exceeds 5%, the system automatically reduces the baud rate by 20% (such as from 115200bps to 92160bps), recalculates the check bit setting (such as changing the original CRC-16 check to a more stringent CRC-32 check), and re-evaluates the communication quality within 1 minute. If it is still abnormal, the protocol is reconfigured to ensure that the communication reliability between the system and the heterogeneous loads is always maintained at a high level.

[0077] Embodiment 3

[0078] After obtaining the equipment type of each heterogeneous load based on the above-mentioned embodiments 1 and 2 and establishing a reliable communication connection, this embodiment also provides a method for intelligent scheduling of energy supply and demand of heterogeneous loads in a building, which includes predicting energy demand based on identification and communication data using a time series algorithm, specifically including building a multivariate time series prediction model based on historical energy consumption data, real-time power demand values ​​of heterogeneous loads, and external environmental factors, and using a deep learning algorithm to predict the energy demand of each heterogeneous load. Then, based on the energy prediction structure and the importance, energy consumption characteristics, and operating status of each heterogeneous load, a dynamic priority allocation algorithm is used to formulate an energy management strategy, such as determining the operating power and start-stop sequence of each heterogeneous load. Real-time monitoring is performed during execution, and the strategy is adjusted in time when the deviation is large or the power grid is abnormal, so as to achieve efficient energy utilization and stable system operation.

[0079] Embodiment 4

[0080] See also Figure 3 As shown, based on the above-mentioned embodiments 1 and 2, an embodiment of the present invention provides an intelligent scheduling system for energy supply and demand of heterogeneous loads in a building, including the following modules:

[0081] Data acquisition module M300: includes a multi-source data acquisition unit and a data preprocessing unit.

[0082] Multi-source data acquisition unit: responsible for collecting multi-type data from heterogeneous load devices widely distributed in zero-carbon park buildings. For electrical parameters, it is equipped with high-precision, wide-range current transformers and voltage transformers, with a measurement accuracy of up to 0.1, covering the common power consumption range of zero-carbon park buildings. For example, the measurable current range is -50A to 50A, the measurable voltage range is 190V to 250V, and the sampling frequency is up to 2000Hz, ensuring that subtle changes and instantaneous fluctuations in electrical parameters can be accurately captured, providing rich raw data for subsequent analysis.

[0083] For devices with communication interfaces, such as smart air conditioners, network charging piles, smart lighting systems, etc., it is equipped with a variety of communication interface adapter cards, supporting RS485, CAN bus, Ethernet and various mainstream brand equipment-specific communication protocol interfaces, which can be fully compatible with the communication connections of devices of different ages and brands, and collect the internal operating status data of the equipment, such as the compressor operating frequency, fan speed, evaporator temperature of the air conditioner, the battery management system data of the charging pile (including battery health status, remaining power prediction, etc.), the dynamic power curve during charging, the brightness adjustment level of smart lighting, lighting area division information, etc. At the same time, each collected data point is given a unique time mark and location tag to accurately trace the data source and time sequence, and realize accurate synchronization and integrated analysis of data.

[0084] Data preprocessing unit: Perform multi-stage preprocessing on the collected raw data. First, use the data filtering algorithm to remove abnormal data points caused by electromagnetic interference, signal transmission noise, etc. For example, use the median filtering algorithm to smooth the instantaneous peaks or valleys in the current and voltage data. Secondly, perform data format standardization conversion to uniformly convert data from different sources and formats into the standard data format specified within the system to facilitate subsequent data processing and storage. Furthermore, perform integrity check and verification on the data, and use the data redundancy check algorithm (such as CRC cyclic redundancy check) to ensure that the data is not lost or tampered with during the collection and transmission process. Re-collect or mark the data that fails the verification to ensure the reliability and accuracy of the data.

[0085] Feature extraction module M301: includes an electrical feature deep extraction unit and an operation mode intelligent analysis unit.

[0086] Electrical feature deep extraction unit: Deep mining based on traditional electrical feature calculations. In addition to calculating basic parameters such as three-phase active power, reactive power, and power factor, the power fluctuation spectrum characteristics are further analyzed, and the power signal is decomposed at multiple scales using wavelet transform to extract power fluctuation characteristics in different frequency bands to distinguish the power change modes of different types of equipment under different operating conditions. For example, for motor equipment, its low-frequency power fluctuation is closely related to the load change of the equipment, while the high-frequency fluctuation may reflect the electromagnetic noise or mechanical vibration of the motor; for electronic equipment, the power fluctuation spectrum shows different distribution characteristics, which are related to the working characteristics of components such as the switching power supply and rectifier circuit inside the equipment. At the same time, the current and voltage waveforms are jointly analyzed, and characteristic indicators such as the phase difference change rate and waveform similarity between current and voltage are calculated to further enrich the electrical feature dimensions and improve the ability to distinguish heterogeneous load equipment.

[0087] Operation mode intelligent analysis unit: Build highly customized operation mode analysis models for different types of heterogeneous load equipment. For central air-conditioning systems, in addition to the basic information such as temperature and operation mode collected, it also combines outdoor meteorological data (such as temperature, humidity, solar radiation intensity, etc.), the internal space layout of the building and the distribution of personnel (obtained through the building management system), and uses machine learning algorithms (such as decision tree algorithms) to build an intelligent operation mode analysis model for central air-conditioning, which can accurately predict the optimal operation mode and parameter settings of air-conditioning under different environmental conditions and personnel needs. For charging piles, based on multi-source data such as battery type, charging current and voltage curve change trends, and vehicle battery management system feedback information, deep learning algorithms (such as long short-term memory network-LSTM) are used to predict the remaining battery charging time and the battery status at the end of charging (such as changes in battery health, whether there is a risk of overcharging, etc.), and intelligently adjust the charging strategy according to the grid load situation and electricity price fluctuation information to achieve dual optimization of charging efficiency and grid stability.

[0088] Dynamic recognition model module M302: includes a hybrid model building and training unit and a model online updating and optimization unit.

[0089] Hybrid model construction and training unit: A hybrid model integrating multiple machine learning algorithms is used to construct a dynamic identification model for heterogeneous loads. With support vector machine (SVM) as the core algorithm, a hierarchical identification model architecture is constructed by combining the advantages of decision tree algorithm and neural network algorithm. In the model training stage, a large amount of unlabeled data is first used to pre-train and reduce the data features using autoencoders to extract the deep feature representation of the data. Then the pre-training results are used as the input of SVM and decision tree algorithms for supervised training. For some complex nonlinear classification problems, neural network algorithms are introduced for local optimization and fine-tuning. For example, for new smart home appliances, their electrical characteristics and operating modes may be quite different from those of traditional devices. Neural network algorithms can quickly capture these new features through adaptive learning capabilities and improve the recognition accuracy of the model for new devices. In the training process, adaptive learning rate adjustment strategies and regularization techniques (such as L1 and L2 regularization) are used to prevent model overfitting. At the same time, the k-fold cross-validation method (k=10) is used to comprehensively evaluate the model and optimize parameters to ensure that the model has good generalization capabilities on different data subsets.

[0090] Model online update and optimization unit: Realize real-time online update and optimization of the model during system operation. When new heterogeneous load equipment is connected or the operating status of existing equipment changes significantly, the model update mechanism is automatically started. First, the new data is preliminarily screened and evaluated to determine its potential impact on the model. For data with significant feature differences, an incremental learning algorithm (such as an incremental learning algorithm based on SVM) is used to gradually integrate it into the existing model and update the decision boundary and parameter settings of the model. At the same time, the model performance is evaluated and optimized regularly, and the model is retrained and parameter adjusted using a large amount of newly accumulated data to maintain a high degree of adaptability between the model and the actual operating environment, ensuring that the recognition accuracy of the model is always maintained at a high level (such as above 95%).

[0091] Adaptive communication configuration module M303: includes a protocol intelligent detection and identification unit and an adaptive communication parameter dynamic adjustment unit.

[0092] Protocol intelligent detection and identification unit: It has powerful protocol intelligent detection and identification capabilities. When heterogeneous load devices with unknown communication protocols are connected to the system, the multi-protocol parallel detection mechanism is automatically started, and detection requests based on multiple mainstream communication protocols (such as Modbus, BACnet, OPC UA, etc.) are sent to the device at the same time, covering different baud rates, data bits, stop bits, checksums and other parameter combinations, and quickly scanning the range of communication protocols that the device may use. At the data receiving end, the protocol feature automatic extraction and analysis technology is used to construct the protocol feature vector based on multiple features such as data frame structure, data content semantics, and communication timing. Then, the protocol feature vector is classified and identified using a deep learning-based protocol classification model (such as convolutional neural network-CNN). It can accurately identify the type of communication protocol used by the device in a very short time (such as a few seconds to tens of seconds). Even in the face of some private protocols or new protocols, it can effectively identify or provide similar protocol references through feature learning and pattern matching capabilities, providing accurate basis for subsequent communication configuration.

[0093] Adaptive communication parameter dynamic adjustment unit: During the communication process, the communication parameters can be dynamically adjusted according to the real-time communication quality and the operating status of the equipment. By real-time monitoring of key indicators such as the signal strength, signal-to-noise ratio, data packet loss rate, and transmission delay of the communication link, a communication parameter adjustment decision model is established using a fuzzy control algorithm. For example, when the data packet loss rate exceeds the set threshold (such as 5%) and the signal strength is weak, the communication baud rate is automatically reduced, the data check bit is increased, and the size and frequency of the transmitted data frame are adjusted to improve the reliability and stability of the communication; when the communication delay is too long and affects the real-time control of the equipment, the transmission layer parameters of the communication protocol are adjusted first (such as using a more efficient transmission protocol or optimizing the caching strategy of the transmission layer), and at the same time, dynamic priority sorting is performed according to the urgency of the equipment and the importance of the data to ensure the timely transmission of important data and the effective control of the equipment. In addition, when the operating status of the equipment changes (such as from low-load operation to high-load operation), the communication parameters are automatically adjusted to adapt to the communication needs of the equipment under different operating conditions, so as to achieve the optimal configuration and utilization of communication resources.

[0094] Embodiment 5

[0095] See also Figure 4 As shown, based on the above-mentioned embodiment 3, an embodiment of the present invention provides an intelligent scheduling system for energy supply and demand of heterogeneous loads in a building. In addition to the modules in embodiment 3, this embodiment also includes:

[0096] Energy management and control module M304: includes energy demand intelligent prediction unit and equipment collaborative optimization control unit.

[0097] Energy demand intelligent prediction unit: Based on historical energy consumption data, real-time heterogeneous load operation data and external environmental factors (such as seasonal changes, weather conditions, electricity price fluctuations, etc.), a multivariate time series prediction model is constructed, and deep learning algorithms (such as recurrent neural network-RNN and its variants LSTM, GRU, etc.) are used to accurately predict the energy demand of various heterogeneous loads in zero-carbon park buildings. For example, for the winter heating season, combined with meteorological data such as outdoor temperature, wind speed, humidity, and factors such as the insulation performance of the building and the activity patterns of personnel, the energy demand of heating equipment in different time periods (such as every hour and every day) is predicted; for the summer cooling season, considering factors such as solar radiation intensity, building orientation, and air conditioning use time distribution, the energy consumption changes of the air conditioning system are predicted. At the same time, energy procurement plans and equipment scheduling strategies are formulated in advance based on the prediction results, such as starting the charging of energy storage equipment in advance or increasing the operation of adjustable loads during the low electricity price period, reducing the operation of unnecessary loads during the peak electricity price period, reducing energy procurement costs and improving energy efficiency.

[0098] Equipment collaborative optimization control unit: realizes collaborative optimization control of heterogeneous load equipment in zero-carbon park buildings. According to the energy demand forecast results and real-time energy supply, the multi-agent system (MAS) theory is used to build an equipment collaborative control framework, and each heterogeneous load equipment is regarded as an intelligent agent. Through information interaction and collaborative decision-making between intelligent agents, load balancing and energy optimization allocation between equipment are realized. For example, during peak hours of electricity consumption, the operation of central air-conditioning systems, electric water heaters, elevators and other equipment is coordinated. According to the importance, energy consumption characteristics, operating status and other factors of each equipment, a dynamic priority allocation algorithm is used to determine the operating power and start-stop sequence of each equipment, giving priority to the normal operation of important equipment (such as elevators, fire-fighting equipment, etc.), while power limiting or time-sharing control of non-critical equipment (such as partial area lighting, adjustable electric heaters, etc.) is performed to avoid power overload and energy waste. In addition, through collaborative control between equipment, the cascade utilization of energy and waste heat recovery are realized, such as using the condensation heat of central air conditioning to provide preheating for the domestic hot water system, improving the comprehensive utilization efficiency of energy and the overall performance of the system.

[0099] Beneficial effects: The present invention provides a method and system for intelligent scheduling of energy supply and demand of heterogeneous loads in zero-carbon campus buildings, which realizes accurate identification of heterogeneous loads through multi-dimensional data collection and feature extraction. For example, in actual test scenarios, the recognition accuracy rate of zero-carbon campus buildings containing air conditioners, charging piles and other electrical equipment of various brands can be as high as 95% or more, which is significantly improved compared to traditional recognition methods that only rely on a single electrical parameter or simple rule judgment. This enables the building energy management system to accurately know the type and operating status of each load, providing a solid foundation for subsequent refined management.

[0100] In terms of adaptive communication protocol configuration, the present invention can automatically detect and identify the communication protocols used by various heterogeneous loads, whether it is the common RS485, Ethernet protocol, or the proprietary protocols of some specific devices, and can quickly adapt. When testing different types of newly connected devices, the average protocol recognition time can be controlled within 30 seconds, and the success rate of communication connection establishment exceeds 98%. This greatly improves the compatibility and scalability of the system, effectively solves the access difficulty problem caused by the mismatch of communication protocols when facing new devices or niche devices in traditional systems, and reduces the cost and difficulty of system integration and maintenance.

[0101] Based on accurate load identification and stable communication connection, the energy management and control module of the present invention can formulate intelligent energy management strategies according to real-time load demand and energy supply conditions. By rationally scheduling the operation of various heterogeneous loads, such as arranging charging of energy storage equipment during low-power consumption periods, increasing power consumption of adjustable loads, and limiting the power of non-critical loads during peak power consumption periods, the overall energy consumption of the building is optimized. It has been verified by actual cases that after implementing the system of the present invention, the energy consumption of zero-carbon park buildings can be reduced by 20%-30%, and the peak-to-valley difference rate can be reduced by 30%-40%, which effectively reduces the power supply pressure of the power grid and improves energy utilization efficiency. At the same time, it also saves users a lot of energy costs, with significant economic and social benefits.

[0102] In addition, the incremental learning mechanism of the present invention enables the system to continuously adapt to the access of new equipment and changes in the operating status of equipment. With the replacement or functional upgrade of equipment in the zero-carbon campus building, the system can automatically learn new load characteristics and communication protocol modes without large-scale manual reconfiguration and model training, ensuring the long-term effectiveness and adaptability of the system, extending the service life of the system, reducing the additional investment caused by technology iteration, and further enhancing the value and competitiveness of the present invention in practical applications.

[0103] The above is only a preferred embodiment of the present invention, and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technician familiar with this profession can make some changes or modifications to equivalent embodiments of equivalent changes using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A method for intelligent scheduling of energy supply and demand of heterogeneous loads in a building, characterized in that: The method comprises the following steps: Acquire different types of heterogeneous load data, wherein the heterogeneous load data includes the equipment type of the load and the corresponding electrical parameters and operation data; Establishing a dynamic identification model for heterogeneous loads, using the heterogeneous load data after feature extraction as a training sample set to train the model, and obtaining an identification model for identifying equipment types; Obtaining heterogeneous load data of the equipment to be detected connected to the building, and determining the type of the equipment to be detected based on the recognition model after feature extraction; Determine the communication protocol type based on the device type and establish a communication connection; Based on identification and communication data, time series algorithms are used to predict energy demand, and energy management strategies are formulated in combination with heterogeneous load data.

2. The intelligent scheduling method for heterogeneous load energy supply and demand in a building as claimed in claim 1 is characterized in that: The feature extraction comprises the following steps: Calculating active power, reactive power, power factor and current waveform characteristics based on the electrical parameters; A feature vector is constructed for the operation data according to the equipment type of the heterogeneous load.

3. The intelligent scheduling method for heterogeneous load energy supply and demand in a building as claimed in claim 1, characterized in that: The step of establishing a dynamic identification model for heterogeneous loads includes: The support vector machine (SVM) algorithm is used to build a dynamic identification model for heterogeneous loads. The collected heterogeneous load data is used as the training sample set in is the feature vector of the i-th sample, y i is the corresponding device type label; The radial basis function is selected as the kernel function to train the SVM model, and the decision function parameters are solved by the sequential minimum optimization algorithm.

4. The intelligent scheduling method for heterogeneous load energy supply and demand in a building as claimed in claim 1, characterized in that: The steps for developing an energy management strategy include: Conduct statistical analysis based on heterogeneous load data from different types of equipment; Determine the energy consumption characteristics of different types of equipment; After determining the type of equipment according to the identification model, synchronously outputting the power demand value and energy consumption characteristics of the equipment; Based on historical energy consumption data, real-time power demand values ​​of heterogeneous loads, and external environmental factors, a multivariate time series prediction model is constructed, and a deep learning algorithm is used to predict the energy demand of each heterogeneous load. According to the energy forecast structure and the importance, energy consumption characteristics and operating status of various heterogeneous loads, a dynamic priority allocation algorithm is used to formulate an energy management strategy.

5. The intelligent scheduling method for energy supply and demand of heterogeneous loads in a building as claimed in claim 1, characterized in that: The method further comprises the following steps: Regularly testing and evaluating the recognition model; If the accuracy of the recognition model is lower than the set value, analyze and determine the cause; Obtain new heterogeneous load data based on the analyzed cause type, extract features and enter them into the training set; Keep the algorithm unchanged and adjust the parameters to retrain the model based on the updated training set.

6. The intelligent scheduling method for heterogeneous load energy supply and demand in a building as claimed in claim 1, characterized in that: The step of determining the type of communication protocol based on the device type and establishing a communication connection includes: Design different communication detection instructions for the device types of heterogeneous loads, wherein the communication detection instructions include operation data request instructions corresponding to the device types, and the instruction formats include multiple communication protocol formats and cover different communication parameters; Detect physical connections and communication signals when devices are connected, and determine identifiability and communication capabilities; If the judgment is passed, a corresponding communication detection instruction is sent to it based on the device type.

7. The intelligent scheduling method for energy supply and demand of heterogeneous loads in a building as claimed in claim 6, characterized in that: The step of determining the type of communication protocol based on the characteristic information of the response data includes: Conducting in-depth analysis on the collected response data to extract protocol feature information; Comparing the extracted protocol feature information with a pre-stored communication protocol feature library, and using a similarity algorithm to calculate the similarity between the protocol feature information and each known protocol feature; Determine the communication protocol type based on the similarity results.

8. The intelligent scheduling method for energy supply and demand of heterogeneous loads in a building as claimed in claim 7, characterized in that: The step of calculating the similarity includes: Determine the weight values ​​of each feature in different protocols; Comparing the protocol feature information extracted from the response data with features of corresponding multiple protocols; The similarity is calculated based on the weights of the same features.

9. An intelligent dispatching system for energy supply and demand of heterogeneous loads in a building, characterized in that: The system comprises: A data acquisition module is used to obtain heterogeneous load data, wherein the heterogeneous load data includes the equipment type of the load and the corresponding electrical parameters and operation data; A feature extraction module, used for extracting features from the heterogeneous load data; A dynamic identification model module is used to establish a dynamic identification model for heterogeneous loads, and train the model using the heterogeneous load data after feature extraction as a training sample set to obtain an identification model for identifying the type of equipment; and determine the type of the equipment to be detected based on the identification model and the heterogeneous load data of the equipment to be retrieved; An adaptive communication configuration module, used to determine the type of communication protocol according to the device type and establish a communication connection; The energy management and control module is used to predict energy demand based on identification and communication data using time series algorithms, and to formulate energy management strategies in combination with heterogeneous load data.

10. The intelligent dispatching system for energy supply and demand of heterogeneous loads in a building as claimed in claim 1, characterized in that: The dynamic identification model is also used to determine the power demand value and energy consumption characteristics of the equipment according to the equipment information; the energy management and control module is specifically used to build a multivariate time series prediction model based on historical energy consumption data, real-time power demand values ​​of heterogeneous loads and external environmental factors, and use a deep learning algorithm to predict the energy demand of each heterogeneous load; It is also used to formulate energy management strategies using a dynamic priority allocation algorithm based on the energy forecast structure and the importance, energy consumption characteristics and operating status of each heterogeneous load.

Images