Electric heating system control method based on source load fluctuation response

Abstract

The invention discloses an electric heating system control method based on source load fluctuation response, which belongs to the technical field of electric heat storage system control, and comprises the following steps: an edge gateway collects and uploads various equipment data; the data center carries out cleaning, alignment and storage; constructing a source and load prediction model based on the cleaning data; the optimization module formulates a control strategy to generate a decision matrix; finally, a control instruction is generated and issued to the equipment, dynamic regulation and control are completed, dynamic cooperative control over all the control equipment in the electric heating system is achieved, the peak clipping and valley filling capacity and the energy utilization efficiency of the whole system are improved, high-precision trend prediction is achieved, the control target is optimized in real time according to the source load prediction result, and the control efficiency is improved. A coordination control instruction of multiple control devices is formulated, various control device protocols are adapted through a uniformly packaged control instruction format, accurate issuing and state feedback of the instruction are realized, and the robustness of the system is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of electric heat storage system control, and particularly relates to an electric heat system control method based on source-load fluctuation response. BACKGROUND

[0002] With the continuous increase of new energy access proportion, the real-time balance between source (power supply side) and load (power consumption side) in the power system becomes increasingly complex.

[0003] Especially in the integrated energy system containing photovoltaic, energy storage, electric heat equipment (such as heat pump, heat storage tank, etc.), affected by light, environmental temperature and load fluctuation, etc., the source-load fluctuation is frequent, and the traditional fixed scheduling strategy has been difficult to meet the safety and energy efficiency optimization requirements of the system.

[0004] In the prior art, a simple time series prediction or rule control method is usually adopted to locally control single devices such as photovoltaic power generation, battery energy storage and heat pump operation. This method has the following disadvantages:

[0005] In terms of data processing, only a single data source or coarse-grained data is used, and there is a lack of standardized processing and effective fusion of multi-source heterogeneous data.

[0006] In terms of prediction calculation, there is a lack of unified modeling of source end (such as photovoltaic, energy storage) and load end (such as heat load demand), the prediction ability is limited, and it is difficult to accurately capture the complex fluctuation rule.

[0007] In terms of control strategy, the control decision is usually based on static setting, and there is a lack of real-time response to dynamic changes of source and load and global optimization, resulting in poor peak clipping and valley filling effect of the system and low energy utilization rate.

[0008] In terms of execution mechanism, the instruction issuing and device protocol conversion lack standardized process, and problems such as instruction delay and error execution are prone to occur, affecting the reliability of the system. SUMMARY

[0009] The purpose of the present application is to provide an electric heat system control method based on source-load fluctuation response, which solves the dynamic cooperative control of various control devices in the electric heat system and improves the peak clipping and valley filling ability and energy utilization efficiency of the overall system.

[0010] To achieve the above purpose, the application adopts the following technical solutions:

[0011] An electric heat system control method based on source-load fluctuation response comprises the following steps:

[0012] Step 1: deploying a plurality of edge gateways in the electric heating system, obtaining raw data of each sensor device in the electric heating system through the edge gateway; the edge gateway encapsulates the raw data in a unified format to generate a raw data packet, and sends the raw data packet to a data center;

[0013] Step 2: establishing a data preprocessing module in the data center, the data preprocessing module verifies, unpacks and aligns the raw data packet to obtain raw data, preprocesses the raw data according to the attributes of the raw data to obtain cleaned raw data, establishes a preprocessed data set, and stores the preprocessed data set in a database;

[0014] Step 3: establishing a model module in the data center, the model module constructs a source prediction model and a load prediction model, uses the source prediction model to predict future photovoltaic power generation trends and battery SOC change trends, uses the load prediction model to predict future heat load change trends, and outputs a multi-dimensional prediction sequence according to the prediction results of the source prediction model and the prediction results of the load prediction model;

[0015] Step 4: establishing an optimization module in the data center, the optimization module obtains the multi-dimensional prediction sequence, constructs an objective function, formulates a control strategy according to the calculation result of the objective function, generates a decision matrix, and the decision matrix includes start-stop instructions or power adjustment instructions of each control device;

[0016] Step 5: establishing an instruction issuing module in the data center, the instruction issuing module generates a control instruction packet in a unified format according to the decision matrix, and distributes the control instruction packet to each edge gateway;

[0017] The edge gateway obtains the control instruction by unpacking the control instruction packet, constructs a protocol instruction according to the communication protocol of each control device, and sends the protocol instruction to the corresponding control device.

[0018] Preferably, the raw data includes photovoltaic power generation power, battery SOC, battery charge and discharge power, heat pump operating state, heat pump output heat power, heat storage tank water temperature, heat storage tank water level, heat load side flow, heat load side water supply temperature, heat load side return water temperature and ambient temperature.

[0019] Preferably, when step 2 is performed, the following steps are specifically included:

[0020] Step 2-1: after the data preprocessing module obtains the raw data packet, the integrity of the raw data packet is verified by CRC verification: if the verification is successful, step 2-2 is performed; otherwise, the raw data packet is discarded, and a processing log is generated;

[0021] Step 2-2: the data preprocessing module splits the key fields in the raw data packet according to the encapsulation format to obtain the to-be-processed data, including type, device ID, timestamp and data segment;

[0022] Steps 2-3: Use interpolation or nearest-neighbor padding to pad the timestamps of the data to be processed;

[0023] Steps 2-4: For the completed data to be processed, preprocess it according to the data type attributes to obtain preprocessed data;

[0024] The preprocessing methods specifically include: applying a moving average filter to the photovoltaic power generation; removing outliers from the battery SOC; truncating the battery power on both sides; detecting abnormal conditions in the heat pump operation; physically limiting the water temperature and level in the thermal storage tank; deburring the flow rate on the heat load side; and verifying the ambient temperature within the normal range.

[0025] Steps 2-5: Standardize the units of the preprocessed data;

[0026] Steps 2-6: Create preprocessed datasets for each preprocessed data and store them in the database.

[0027] Preferably, step 3 includes the following steps:

[0028] Step 3-1: Construct the source prediction model, which includes a photovoltaic output prediction model and a battery SOC evolution model:

[0029] The calculation formula for the photovoltaic output prediction model is as follows:

[0030] Ppv(t+Δt)=α×Ppv(t)+β×Tenv(t)+γ;

[0031] Where Ppv(t+Δt) is the predicted photovoltaic output power at a future time, t is the time, Δt is the time increment, Ppv(t) is the current photovoltaic power, Tenv(t) is the current ambient temperature, and α, β, and γ are all regression coefficients;

[0032] The calculation formula for the battery SOC evolution model is as follows:

[0033]

[0034] Where Pbat(t) is the current battery charging and discharging power; η is the charging and discharging efficiency; Cbat is the total battery capacity; SOC(t) is the current battery SOC; and SOC(t+Δt) is the predicted battery SOC at a future time.

[0035] Step 3-2: Construct the load prediction model. The calculation formula for the load prediction model is as follows:

[0036] Qload(t+Δt) = a x Flow(t) + b x (Tin(t) - Tout(t)) + c;

[0037] Wherein, Qload(t+Δt) is the linear heat load prediction result, Flow(t) is the heat load side flow, Tin(t) and Tout(t) are the heat load side water supply temperature and heat load side return water temperature respectively, a, b and c are regression coefficients;

[0038] Step 3-3: According to the prediction result of the source prediction model and the prediction result of the load prediction model, output the future multi-dimensional prediction sequence, and the multi-dimensional prediction sequence includes the predicted photovoltaic power generation, the predicted battery SOC and the predicted load demand.

[0039] Preferably, when step 4 is executed, it specifically includes the following steps:

[0040] Step 4-1: Construct a target function, and the target of the target function is peak shaving and valley filling, and the specific calculation formula is as follows:

[0041]

[0042] Pnet(t) = Pload(t) + Ppv(t) - Pbat(t);

[0043] Wherein, is the target smooth power level;

[0044] Step 4-2: Obtain a control device list and develop a control strategy, and the control device list includes a battery charge and discharge controller, a heat pump start and stop controller and a heat pump power adjustment controller;

[0045] The control strategy includes: when the predicted photovoltaic is excessive, preferentially charge, start the heat pump to increase the heat load; when the predicted load peak is high, the battery is discharged, and part of the load is closed; when the water temperature of the heat storage tank reaches the upper limit, the heat pump is closed to prevent the heat pump from being overcharged;

[0046] Step 4-3: According to the control strategy, generate control instructions of each control device, and integrate into a decision matrix.

[0047] Preferably, when step 5 is executed, the unified format control instruction package generated by the data center specifically includes: [header instruction]+[target device ID]+[control action type]+[parameter]+[timestamp]+[CRC]+[tail instruction].

[0048] The source-load fluctuation response-based electric heating system control method provided by the application solves the dynamic collaborative control of various control devices in the electric heating system, improves the peak clipping and valley filling capacity and energy utilization efficiency of the overall system, and has the technical problems that the application uniformly collects multi-dimensional original data such as photovoltaic power, energy storage state, heat pump operation, heat load demand, and environmental conditions, performs standardization cleaning and synchronization, establishes a prediction model of photovoltaic output, energy storage evolution, and heat load change respectively in combination with source and load characteristics, realizes high-precision trend prediction, optimizes the control target in real time according to the source-load prediction result, formulates the coordinated control instruction of multiple control devices, adapts to various control device protocols through the unified encapsulated control instruction format, realizes accurate issuance and state feedback of the instruction, and improves the system robustness. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 is the main flowchart of the application;

[0050] Figure 2 is the flowchart of step 2 of the application;

[0051] Figure 3 is the flowchart of step 3 of the application;

[0052] Figure 4 is the flowchart of step 4 of the application. DETAILED DESCRIPTION

[0053] The source-load fluctuation response-based electric heating system control method shown in Figures 1-4 includes the following steps:

[0054] Step 1: Deploy multiple edge gateways in the electric heating system, obtain the original data of various sensor devices in the electric heating system through the edge gateways, and generate original data packets after the edge gateways perform unified format encapsulation on the original data, and send the original data packets to the data center.

[0055] The original data includes photovoltaic power generation power, battery SOC, battery charging and discharging power, heat pump operation state, heat pump output heat power, heat storage tank water temperature, heat storage tank water level, heat load side flow, heat load side water supply temperature, heat load side return water temperature, and environmental air temperature.

[0056] In this embodiment, the edge gateway can adopt an embedded micro industrial computer or an AI edge box and the like hardware device, and a C++ / Python service is arranged inside the edge gateway, which is used to perform tasks such as loading of various protocol drivers (Modbus RTU / TCP, MQTT, HTTP), real-time acquisition of original data, format encapsulation and packaging, and uploading to the center in the MQTT or HTTP mode.

[0057] The specific data sources of the original data are shown in Table 1 as follows:

[0058]

[0059] Table 1

[0060] Each edge gateway corresponds to a deployment location, and the same edge gateway can collect raw data of multiple sensors or transmitters in its deployment location according to the field situation. The format of the raw data is in the Modbus protocol format, which is as follows:

[0061] [header instruction]+[data type identifier]+[device ID]+[timestamp]+[data value]+[CRC check]+[tail instruction];

[0062] [Data type identifier] represents the attribute of the data, such as 01 = photovoltaic power, 02 = SOC, and 03 = water temperature.

[0063] [Device ID] is the unique identifier of the sensor / transmitter.

[0064] [Data value] is a floating point / fixed point data representing the specific data value collected by each sensor.

[0065] In this embodiment, the collection frequency of the edge gateway is controlled to be collected once every 5 seconds / 10 seconds, and after encapsulating multiple raw data, a unified data packet is formed and uploaded to the data center. The encapsulated raw data packet is as follows:

[0066] [Encapsulation header instruction]+[packet number]+[timestamp]+[data number]+{[data segment 1],[data segment 2],...,[data segment N]}+[CRC check]+[encapsulation tail instruction];

[0067] [Packet number] uses incremental numbering to track the order of the packet; [Data number] represents the number of data segments in the packet; [Data segment 1]...[Data segment N] is a single raw data.

[0068] Step 2: Establish a data preprocessing module in the data center. The data preprocessing module performs verification, unpacking, and cache alignment on the raw data packet to obtain raw data, pre-processes the raw data according to its attributes to obtain cleaned raw data, establishes a pre-processed data set, and stores it in a database;

[0069] When performing step 2, the following steps are included:

[0070] Step 2-1: After the data preprocessing module obtains the raw data packet, it verifies the integrity of the raw data packet through CRC check: if the verification is successful, step 2-2 is performed; otherwise, the raw data packet is discarded and a processing log is generated.

[0071] Step 2-2: The data preprocessing module splits the key fields in the original data packet according to the packaging format to obtain the to-be-processed data, including type, device ID, timestamp, and data segment;

[0072] Step 2-3: The interpolation method or the nearest value filling method is used to fill the timestamp of the to-be-processed data.

[0073] In this embodiment, the interpolation method or the nearest value filling method is used to align the data, which solves the problem that in the actual collection process, due to device communication delay, network packet loss, inconsistent collection frequency of different devices, etc., there may be a phenomenon that the timestamps of some key data in the data packet uploaded by the edge device are not continuous or missing.

[0074] The specific method of the interpolation method is: when the time interval of two adjacent data points is within a reasonable range, but there are several missing time points in the middle, the interpolation method is used to estimate the missing time points by using the previous and subsequent valid data points to generate fitting data.

[0075] The specific method of the nearest value filling method is: when the device data is not updated in time, or the sampling frequency is lower than the requirement of the unified time axis, the last valid value is used to fill the subsequent missing time points.

[0076] In this embodiment, the time alignment is based on a unified sampling period, such as 5 seconds or 10 seconds, and the data of all devices will be mapped to this period.

[0077] Step 2-4: For the to-be-processed data after filling, the data is preprocessed according to the type attribute to obtain preprocessed data.

[0078] The preprocessing method specifically includes: sliding mean filter processing of photovoltaic power generation; outlier elimination processing of battery SOC; bilateral truncation processing of battery power; abnormal state detection processing of heat pump operating state; physical range limitation processing of heat storage tank water temperature and heat storage tank water level; deburring processing of heat load side flow; normal range verification processing of environmental temperature;

[0079] In this embodiment, the sliding mean filter method is to smooth the historical power data in a time window, weaken the instantaneous peak, extract the stable trend, and improve the controllability and predictability of the power signal.

[0080] Outlier elimination processing is to detect abnormal points in the data sequence that do not conform to the change rate or boundary rules, identify them as outliers and eliminate them, and retain the credible data sequence.

[0081] Bilateral clipping process is to limit the data within the legal interval by setting upper and lower boundaries such as -30 kW to +30 kW, to prevent false triggering of abnormal state of charge or discharge response instructions, while avoiding misleading data to scheduling logic.

[0082] The principle of abnormal state detection process is to analyze the frequency, duration and other dimensions of the change of state value, identify and filter non-physical abnormal state, and improve the stability of control logic.

[0083] Physical range limitation process is to determine sensor failure or communication interference if the sensor feedback value exceeds the preset range, and to exclude or replace illegal values with safe default values by setting upper and lower threshold values.

[0084] Deburring process is to identify discontinuous jumps and replace them with stable values before and after by judging the instantaneous change rate or local deviation degree, so as to restore the true flow level and improve the accuracy of heat load identification.

[0085] Normal range verification process is to define reasonable upper and lower limits of air temperature (e.g. -30℃ to 60°C), filter out out-of-limit values and mark them as abnormal data, which helps to avoid the impact of false weather judgment on heat pump control strategy.

[0086] Step 2-5: unify the units of the preprocessed data;

[0087] Step 2-6: establish a preprocessed data set for each preprocessed data and store it in the database.

[0088] In this embodiment, structured database (such as PostgreSQL, MySQL) or time series database (such as InfluxDB, TimescaleDB) is used for storage, and the data table structure is shown in Table 2 as follows:

[0089]

[0090] Table 2

[0091] Step 3: establish a model module in the data center, the model module constructs a source prediction model and a load prediction model, uses the source prediction model to predict the future trend of photovoltaic power generation and the trend of battery SOC change, uses the load prediction model to predict the future trend of heat load change, and outputs the future multi-dimensional prediction sequence according to the prediction results of the source prediction model and the prediction results of the load prediction model;

[0092] When performing step 3, the following steps are included:

[0093] Step 3-1: construct a source prediction model, the source prediction model includes a photovoltaic output prediction model and a battery SOC evolution model:

[0094] The calculation formula of the photovoltaic output prediction model is as follows:

[0095] Ppv(t+Δt)=α×Ppv(t)+β×Tenv(t)+γ;

[0096] wherein, Ppv(t+Δt) is the predicted photovoltaic output power at future time t+Δt, t is the time, Δt is the time step, Ppv(t) is the current photovoltaic power, i.e. the photovoltaic power generation power, Tenv(t) is the current environmental temperature, α, β, γ are all regression coefficients, and α, β, γ reflect the change relationship of the photovoltaic pv output with the power and the air temperature.

[0097] The calculation formula of the battery SOC evolution model is as follows:

[0098]

[0099] wherein, Pbat(t) is the current battery charging and discharging power, the discharging is positive, and the charging is negative; η is the charging and discharging efficiency, and the value is between 0.9 and 0.95; Cbat is the total capacity of the battery, SOC(t) is the current battery SOC, and SOC(t+Δt) is the predicted battery SOC at future time;

[0100] Step 3-2: constructing a load prediction model, and the calculation formula of the load prediction model is as follows:

[0101] Qload(t+Δt)=a×Flow(t)+b×(Tin(t)-Tout(t))+c;

[0102] wherein, Qload(t+Δt) is the linear heat load prediction result, Flow(t) is the heat load side flow, Tin(t) and Tout(t) are respectively the heat load side water supply temperature and the heat load side return water temperature, and a, b, c are all regression coefficients and are obtained by fitting the historical heat load behaviors;

[0103] Step 3-3: outputting the future multi-dimensional prediction sequence according to the prediction result of the source prediction model and the prediction result of the load prediction model, and the multi-dimensional prediction sequence includes the predicted photovoltaic power generation, the predicted battery SOC and the predicted load demand.

[0104] In the embodiment, the source prediction model and the load prediction model are used to predict a plurality of future time points, and the prediction values are outputted, and the outputted prediction values are combined into a multi-dimensional prediction sequence, for example, predicting once every 10 minutes within the next 60 minutes, and the multi-dimensional prediction sequence is shown in Table 3:

[0105]

[0106] Table 3

[0107] In this embodiment, the heat pump operating state is used to predict the state (start / stop), mainly to provide a decision boundary for the scheduling optimization model to avoid frequent start-stop; the heat pump output heat power is used as the upper limit of the heat source output capacity, to calculate the heat pump working efficiency, and to assist in determining whether to switch to the heat release mode; the heat storage tank water temperature is used to determine whether the heat pump start-stop can be delayed, and compared with the load to determine whether to release heat or store heat; the heat storage tank water level is used to determine whether the heat storage tank is "full" or "empty", to determine the execution boundary of the control strategy.

[0108] Step 4: Establish an optimization module in the data center, the optimization module obtains a multi-dimensional prediction sequence, constructs a target function, formulates a control strategy according to the calculation result of the target function, and generates a decision matrix, the decision matrix including start-stop instructions or power adjustment instructions of each control device;

[0109] When step 4 is executed, the following steps are specifically included:

[0110] Step 4-1: Construct a target function, the target of the target function being peak shaving and valley filling, and the specific calculation formula being as follows:

[0111]

[0112] Pnet(t)=Pload(t)+Ppv(t)-Pbat(t);

[0113] Wherein, is the target smooth power level; Pnet(t) is the total compliance, and T is a preset time threshold.

[0114] In this embodiment, the target function aims to achieve peak shaving and valley filling, to smooth the power curve of the system, to reduce power fluctuations, and to improve energy efficiency.

[0115] In this embodiment, a constraint function is established according to user requirements, for example:

[0116] Battery charge and discharge power limit:

[0117] min Pbat≤Pbat(t)≤max Pbat;

[0118] SOC limit:

[0119] min SOC≤SOC(t)≤max SOC;

[0120] Step 4-2: Obtain a control device list, formulate a control strategy, and the control device list including a battery charge and discharge controller, a heat pump start-stop controller, and a heat pump power adjustment controller;

[0121] The control strategy includes: when the photovoltaic is predicted to be excessive, preferentially charging and starting the heat pump to increase the heat load; when the load is predicted to be high, discharging the battery and controlling the heat pump power; when the water temperature of the heat storage tank reaches the upper limit, the heat pump is turned off to prevent overcharging of the heat pump;

[0122] In this embodiment, an auxiliary strategy is also set: when the water temperature of the heat storage tank is low, the heat pump power is started or increased; when the heat pump is predicted to be in a state of "high frequency start-stop", frequent switching is avoided by constraint (such as adding a start-stop minimum interval Δt_min) to protect the equipment; the state of the water level of the heat storage tank being close to full / empty is set as a boundary constraint condition (such as limiting the maximum and minimum values of the water temperature and the maximum and minimum values of the water level, thereby limiting the heat pump to continue heating or limiting the heat release condition), which affects the heat pump and heat release strategies.

[0123] Step 4-3: According to the control strategy, control instructions of each control device are generated, and a decision matrix is integrated.

[0124] In this embodiment, the decision matrix is shown in Table 4 as follows:

[0125]

[0126] Table 4

[0127] Step 5: An instruction issuing module is established in the data center, the instruction issuing module generates a control instruction package in a unified format according to the decision matrix, and the control instruction package is distributed to each edge gateway;

[0128] After the edge gateway unpacks the control instruction package, the control instruction is obtained, a protocol instruction is constructed according to the communication protocol of each control device, and the protocol instruction is sent to the corresponding control device.

[0129] When step 5 is executed, the control instruction package in a unified format generated by the data center specifically includes: [header instruction]+[target device ID]+[control action type]+[parameter]+[timestamp]+[CRC]+[tail instruction].

[0130] In this embodiment, the field description of the control instruction package is as follows:

[0131] [Header instruction]: fixed identification, 1 byte, such as 0xAA;

[0132] [Target device ID]: 2-byte unique number of the device, such as 0x01 0x10 (heat pump), 0x02 0x20 (battery controller);

[0133] [Control action type]: 1-byte action type code, such as: 0x01: start the device; 0x02: turn off the device; 0x03: set the power; 0x04: set the temperature threshold;

[0134] [Parameter]: 1-4 bytes, determined according to the action type (such as 2 bytes for setting power, unit 0.1 kW)

[0135] [CRC]: 2 bytes of CRC check;

[0136] [Tail instruction]: fixed identifier, 1 byte, such as 0x55.

[0137] For example: AA 01 10 03 00 28 66F0 A2 50AB 9E 55;

[0138] Among them, 0x01 0x10 is heat pump 1, 0x03 is to set the power, 0x00 0x28 is the parameter, 0x66 0xF0 0xA2 0x50 is the timestamp, 0xAB 0x9E is the CRC, and 0x55 is the tail instruction.

[0139] The electric heating system control method based on source-load fluctuation response provided by the application solves the dynamic collaborative control of various control devices in the electric heating system, improves the peak clipping and valley filling capacity and energy utilization efficiency of the overall system, and solves the technical problem of the application. The application uniformly collects multi-dimensional original data such as photovoltaic power, energy storage state, heat pump operation, heat load demand and environmental conditions, performs standardization cleaning and synchronization, combines the characteristics of the source end and the load end, respectively establishes prediction models of photovoltaic output, energy storage evolution and heat load change, realizes high-precision trend prediction, optimizes the control target in real time according to the source-load prediction result, formulates the coordinated control instructions of multiple control devices, adapts various control device protocols through the unified encapsulated control instruction format, realizes accurate issuing and state feedback of the instructions, and improves the system robustness.

Claims

1. A method for controlling an electric heating system based on source load fluctuation response, characterized in that: The method comprises the following steps: Step 1: deploying a plurality of edge gateways in the electric heating system, and obtaining original data of various sensor devices in the electric heating system through the edge gateways; The edge gateways encapsulate the original data in a unified format to generate original data packets and send the original data packets to a data center; Step 2: establishing a data preprocessing module in the data center, the data preprocessing module verifying, unpacking and aligning the original data packets to obtain original data, preprocessing the original data according to the attributes of the original data to obtain cleaned original data, establishing a preprocessed data set and storing the preprocessed data set in a database; Step 3: establishing a model module in the data center, the model module constructing a source prediction model and a load prediction model, the source prediction model being used to predict future photovoltaic power generation trends and battery SOC change trends, the load prediction model being used to predict future heat load change trends, and a multi-dimensional prediction sequence being output according to the prediction results of the source prediction model and the prediction results of the load prediction model; Step 4: establishing an optimization module in the data center, the optimization module obtaining the multi-dimensional prediction sequence, constructing a target function, formulating a control strategy according to the calculation result of the target function, and generating a decision matrix, the decision matrix comprising start-stop instructions or power adjustment instructions of various control devices; Step 5: establishing an instruction issuing module in the data center, the instruction issuing module generating a control instruction packet in a unified format according to the decision matrix and distributing the control instruction packet to various edge gateways; The edge gateways unpack the control instruction packet to obtain control instructions, construct protocol instructions according to the communication protocols of various control devices, and send the protocol instructions to corresponding control devices.

2. A method of controlling an electric heating system based on source load wave response as claimed in claim 1, characterized in that: The original data comprises photovoltaic power generation power, battery SOC, battery charge-discharge power, heat pump operating state, heat pump output heat power, heat storage tank water temperature, heat storage tank water level, heat load side flow, heat load side water supply temperature, heat load side return water temperature and ambient air temperature.

3. A method of controlling an electric heating system based on source load wave response as claimed in claim 2, characterized in that: When step 2 is performed, the following steps are specifically included: Step 2-1: after the data preprocessing module obtains the original data packet, the integrity of the original data packet is verified by CRC verification: if the verification is successful, step 2-2 is performed; otherwise, the original data packet is discarded and a processing log is generated; Step 2-2: the data preprocessing module splits the key fields in the original data packet according to the encapsulation format to obtain to-be-processed data, including type, device ID, timestamp and data segment; Step 2-3: the interpolation method or the nearest value filling method is used to supplement the timestamp of the to-be-processed data; Step 2-4: the to-be-processed data after the supplement is preprocessed according to the type attribute of the data to obtain preprocessed data; The preprocessing method specifically comprises: performing sliding mean filter processing on the photovoltaic power generation power; performing outlier elimination processing on the battery SOC; performing double-sided truncation processing on the battery power; performing abnormal state detection processing on the heat pump operating state; performing physical range limitation processing on the heat storage tank water temperature and the heat storage tank water level; performing deburring processing on the heat load side flow; and performing normal range verification processing on the ambient air temperature; Step 2-5: the units of the preprocessed data are unified; Step 2-6: Establish the pretreatment data set of each pretreatment data and store it in the database.

4. A method of controlling an electric heating system based on source load wave response as claimed in claim 2, characterized by: When performing step 3, the following steps are specifically included: Step 3-1: Construct a source prediction model, which includes a photovoltaic output prediction model and a battery SOC evolution model: The calculation formula of the photovoltaic output prediction model is as follows: Ppv(t+Δt)=α×Ppv(t)+β×Tenv(t)+γ; Wherein, Ppv(t+Δt) is the predicted photovoltaic output power at future time, t is the time, Δt is the time increment, Ppv(t) is the current photovoltaic power, Tenv(t) is the current environmental temperature, α, β, γ are regression coefficients; The calculation formula of the battery SOC evolution model is as follows: Wherein, Pbat(t) is the current battery charging and discharging power; η is the charging and discharging efficiency; Cbat is the total capacity of the battery, SOC(t) is the current battery SOC, SOC(t+Δt) is the predicted battery SOC at future time; Step 3-2: Construct a load prediction model, and the calculation formula of the load prediction model is as follows: Qload(t+Δt)=a×Flow(t)+b×(Tin(t)-Tout(t))+c; Wherein, Qload(t+Δt) is the linear heat load prediction result, Flow(t) is the heat load side flow, Tin(t) and Tout(t) are the heat load side water supply temperature and heat load side return water temperature respectively, a, b, c are regression coefficients; Step 3-3: According to the prediction results of the source prediction model and the prediction results of the load prediction model, output the future multi-dimensional prediction sequence, which includes the predicted photovoltaic power generation, the predicted battery SOC and the predicted load demand.

5. A method of controlling an electric heating system based on source load wave response as claimed in claim 4, characterized by: When performing step 4, the following steps are specifically included: Step 4-1: Construct a target function, and the target of the target function is peak shaving, and the specific calculation formula is as follows: wherein, target smoothed power level; Step 4-2: Obtain a control device list, develop a control strategy, and the control device list includes a battery charging and discharging controller, a heat pump start-stop controller and a heat pump power adjustment controller; The control strategy includes: when the predicted photovoltaic is excessive, preferentially charge, start the heat pump to increase the heat load; when the predicted load peak is high, the battery is discharged, and part of the load is closed; when the water temperature of the heat storage tank reaches the upper limit, the heat pump is closed to prevent the heat pump from being overcharged; Step 4-3: According to the control strategy, generate control instructions of each control device, and integrate them into a decision matrix.

6. A method of controlling an electric heating system based on source load wave response as claimed in claim 4, characterized by: When performing step 5, the unified format control instruction package generated by the data center specifically includes: [header instruction]+[target device ID]+[control action type]+[parameter]+[timestamp]+[CRC]+[tail instruction].

Images