A heat exchange station load prediction and analysis system based on deep learning

Abstract

The present invention belongs to the technical field of load forecasting, and specifically discloses a heat exchange station load forecasting and analysis system based on deep learning, including an indoor temperature retrieval module, a temperature setting information extraction module, a synchronous indoor temperature extraction module, a set temperature tendency correlation module, a set time tendency correlation module, a heating load information retrieval module, a heating temperature difference tendency correlation module, and a heating load forecasting module. By retrieving the temperature setting records of users during the heating period, extracting the set time from them, identifying the preferred set time of users, and further retrieving the indoor temperature data at these preferred set times, the optimization of the temperature object and the selection range in the meteorological factors affecting load forecasting is realized. On this basis, the load forecasting is more reasonable and accurate, and at the same time, the problems of a large amount of meteorological data occupying storage space and reducing the forecasting efficiency are avoided.

Description

Technical Field

[0001] The present invention belongs to the technical field of load forecasting, and specifically discloses a heat exchange station load forecasting and analysis system based on deep learning. Background Art

[0002] A heat exchange station is a facility used to convert high-temperature hot water or steam generated by a heat source (such as a boiler, a thermal power plant, etc.) into low-temperature hot water suitable for building heating. It transfers heat to the building's heating system through a heat exchanger, thereby achieving efficient heat transfer and distribution. The heat exchange station generates a heating load during the building heating process, that is, the total amount of heat provided to the building within a certain period of time. By accurately predicting the heating load of the heat exchange station, dynamic adjustment of the operating parameters of the heat exchange station can be realized, energy waste can be reduced, and the energy efficiency of the system and the comfort of users can be improved.

[0003] In the prior art, the publication number CN114118862A discloses a heat exchange station dynamic heat load forecasting and control method and system. By obtaining meteorological data and heat exchange station operation data of the area to be processed, where the meteorological data includes outdoor temperature, outdoor humidity, wind speed, etc., and the heat exchange station operation data includes operation duration, operation target temperature, etc., and constructing and dividing a meteorological data set after preprocessing, then constructing a heat exchange station dynamic heat load forecasting model, and further training the forecasting model using the constructed data set, and evaluating the operation effect of the model and making feedback adjustments to the operation state of the heat exchange station to achieve dynamic balance between the supplied heat and the demanded heat. This solution lacks effective screening of meteorological data in terms of time when obtaining meteorological data. The essence of the heat exchange station heating the building is to heat the users inside the building, and the activities of users inside the building are not all-day long. Therefore, not all meteorological data in all time periods are necessary. In this case, using all meteorological data without screening may, on the one hand, introduce noise with irrelevant meteorological data, making the time pattern of user activities not match the time resolution of meteorological data and affecting the prediction accuracy; on the other hand, a large amount of meteorological data will occupy more storage space, increase the cost of data management and storage, and at the same time, it is also easy to increase the burden of data processing and analysis and affect the prediction efficiency.

[0004] In addition, when heating inside a building, the indoor temperature is actually raised to the target temperature, rather than raising the outdoor temperature to the target temperature. There is a certain gap between the indoor temperature and the outdoor temperature, which is affected by the heat insulation effect of the building. Generally speaking, the worse the heat insulation effect of the building, the closer the indoor temperature is to the outdoor temperature. Therefore, the above solution has certain limitations in relying on the outdoor temperature for load forecasting, affecting the accuracy of the prediction results. Summary of the Invention

[0005] To solve the above technical problems or at least partially solve the above technical problems, the present application provides a heat exchange station load prediction and analysis system based on deep learning. Since the most important meteorological factor affecting the load is temperature, the temperature object and the selection range are optimized during the heat exchange station load prediction process, effectively improving the accuracy and rationality of the load prediction.

[0006] The object of the present invention can be achieved by the following technical solutions: A heat exchange station load prediction and analysis system based on deep learning, comprising: an indoor temperature retrieval module, configured to extract the real-time collected indoor temperature from the building temperature control center during the heating period to generate an indoor temperature time series.

[0007] A temperature setting information extraction module, configured to retrieve the temperature setting records of users from the building temperature control center during the heating period, and extract the setting time, setting duration, and setting temperature therefrom.

[0008] A synchronous indoor temperature extraction module, configured to extract the indoor temperature corresponding to the setting time from the indoor temperature time series based on the setting time as the synchronous indoor temperature.

[0009] A setting temperature tendency correlation module, configured to map the temperature setting records of users to the synchronous indoor temperature, generate an indoor temperature classification set by classifying the temperature setting records corresponding to the same synchronous indoor temperature, and then analyze the tendency correlation between the indoor temperature and the setting temperature accordingly.

[0010] A setting time tendency correlation module, configured to classify the retrieved temperature setting records according to the same setting time to generate a setting time classification set, and analyze the preferred setting time and the tendency correlation between the preferred setting time and the setting duration accordingly.

[0011] A heating load information retrieval module, configured to retrieve the heating operation records related to the temperature setting records from the heat exchange station heating control center, and extract the unit heating load therefrom.

[0012] A heating temperature difference tendency correlation module, configured to calculate the heating temperature difference by using the retrieved temperature setting records, and then analyze the tendency correlation between the heating temperature difference and the unit heating load.

[0013] A heating load prediction module, configured to perform heating load prediction by comprehensively considering the tendency correlation between the indoor temperature and the setting temperature, the preferred setting time and the tendency correlation between the preferred setting time and the setting duration, and the tendency correlation between the heating temperature difference and the unit heating load.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The present invention retrieves the temperature setting records of users during the heating period, extracts the set times from them to identify the preferred set times of users, and further retrieves the indoor temperature data at these preferred set times, realizing the optimization of the temperature object and the selection range in the meteorological factors affecting load prediction. On this basis, the load prediction can focus on the meteorological data during the user activity period, ensure that the heating demand during these time periods is met, make the prediction results more reasonable and accurate, and at the same time avoid the problems of a large amount of meteorological data occupying storage space and reducing prediction efficiency.

[0015] 2. When the present invention performs load prediction, it takes into account that the target temperature may not be fixed, that is, the required target temperature of users varies at different indoor temperatures and has a certain correlation with the indoor temperature. Therefore, by retrieving the temperature setting records of users and their corresponding indoor temperatures, the tendency correlation analysis between the indoor temperature and the set temperature is carried out. Furthermore, during load prediction, the tendency target temperature can be obtained for the current indoor temperature, so as to perform more accurate and targeted load prediction, making the load prediction closer to the actual needs of users. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0017] Figure 1 It is a schematic diagram of the connection of each module of the system of the present invention.

[0018] Figure 2 It is a schematic diagram of the storage of the building temperature control center in the present invention for the indoor temperature and temperature setting records.

[0019] Figure 3 It is a flow chart of the tendency correlation analysis between the indoor temperature and the set temperature in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0021] Refer to Figure 1As shown in the figure, a heat exchange station load prediction and analysis system based on deep learning includes an indoor temperature retrieval module, a temperature setting information extraction module, a synchronous indoor temperature extraction module, a set temperature trend correlation module, a set time trend correlation module, a heating load information retrieval module, a heating temperature difference trend correlation module, and a heating load prediction module.

[0022] Among the above, both the indoor temperature retrieval module and the temperature setting information extraction module are connected to the synchronous indoor temperature extraction module. The temperature setting information extraction module and the synchronous indoor temperature extraction module are both connected to the set temperature trend correlation module. The temperature setting information extraction module is connected to the set time trend correlation module. The temperature setting information extraction module is connected to the heating load information retrieval module. The temperature setting information extraction module and the heating load information retrieval module are both connected to the heating temperature difference trend correlation module. The set temperature trend correlation module, the set time trend correlation module, and the heating temperature difference trend correlation module are all connected to the heating load prediction module.

[0023] The indoor temperature retrieval module is used to extract the real-time collected indoor temperature from the building temperature control center during the heating period to generate an indoor temperature time series.

[0024] It should be added that since the heating of the building by the heat exchange station is only limited to winter and not continuous throughout the year, there is a specific heating period, that is, the winter continuous period. Due to different climate types in different regions, the winter continuous periods in each region also vary. For example, for subtropical monsoon climate regions, the winter continuous period is usually from December to February of the following year. In addition, more indoor temperature data are provided for load prediction. The above-mentioned heating period is not limited to the heating period of the current year, and the heating period data of several years away from the current year can be selected. This can increase the diversity and representativeness of the data, thereby improving the accuracy and reliability of the load prediction.

[0025] It should be further added that the collection of indoor temperature is realized by using the intelligent thermostat in the building temperature control center. This is because the intelligent thermostat is equipped with a temperature sensor, and the temperature sensor collects and records temperature data at a high frequency (such as every minute) to ensure the integrity and accuracy of the data. Moreover, the intelligent thermostat is equipped with a storage unit, and the indoor temperature collected by the temperature sensor can be locally stored in the storage unit in the form of collection time + indoor temperature, which is convenient for retrieval.

[0026] The temperature setting information extraction module is used to retrieve the user's temperature setting records from the building temperature control center during the heating period and extract the set time, set duration, and set temperature therefrom.

[0027] It should be added that refer to Figure 2As shown, intelligent thermostats in building temperature control centers are usually equipped with touchscreens or physical buttons through which users can set temperatures. Temperature setting records are generated and stored locally. Additionally, when local storage is limited, the intelligent thermostat can transmit data to a cloud server via wireless communication technologies such as 4G / 5G mobile communication, Wi-Fi, and Bluetooth. The cloud server stores a large amount of historical temperature setting data and indoor temperature data, ensuring not only the security and integrity of the data but also providing the convenience of remote management and analysis.

[0028] The synchronization indoor temperature extraction module is used to extract the indoor temperature corresponding to the temperature setting time from the indoor temperature time series data based on the set time as the synchronization indoor temperature.

[0029] In the present invention, by matching the set time in the temperature setting record with the acquisition time of the indoor temperature, the indoor temperature collected at the set time is screened out to synchronize the timestamps of the temperature setting record and the indoor temperature record, ensuring their temporal correspondence.

[0030] It should be noted that the set time in the temperature setting record reflects the heating demand time of the user in the building. Since the user only generates a heating demand after entering the building, by using the set time in the temperature setting record, the indoor temperature data synchronized with the corresponding time can be screened out. This can accurately reflect the indoor temperature at the heating demand time. By screening through time synchronization, it is ensured that the indoor temperature data used for load prediction is the data at the user's heating demand time, avoiding using data during non-demand times, reducing noise and interference, and improving the accuracy of the prediction.

[0031] See Figure 3 As shown, the set temperature tendency correlation module is used to map the user's temperature setting record with the synchronization indoor temperature. By classifying the temperature setting records corresponding to the same synchronization indoor temperature, an indoor temperature classification set is generated, where the indoor temperature classification set is a set of temperature setting records corresponding to each indoor temperature. Then, the tendency correlation between the indoor temperature and the set temperature is analyzed based on this. The specific analysis is as follows: The set temperature is extracted from each temperature setting record included in the indoor temperature classification set.

[0032] Compare several set temperatures corresponding to the same indoor temperature to identify whether there is a mode. If there is a mode, use the mode as the tendency set temperature corresponding to the indoor temperature. If there is no mode, calculate the range of several set temperatures corresponding to the same indoor temperature and compare it with the configured allowable range. Exemplarily, the allowable range is 0.1 °C. If the calculated range value is less than or equal to the allowable range, take the average value of several set temperatures corresponding to the same indoor temperature as the tendency set temperature corresponding to the indoor temperature. Otherwise, take the maximum value of several set temperatures corresponding to the same indoor temperature as the tendency set temperature corresponding to the indoor temperature.

[0033] It should be noted that after synchronizing the temperature setting records with the indoor temperature, considering that in some cases, the same indoor temperature may correspond to multiple different temperature setting records, that is, the same indoor temperature corresponds to multiple set temperatures. This one-to-many relationship will lead to data redundancy, affecting the accuracy and efficiency of load forecasting. Through tendency analysis, the most representative set temperature can be identified and selected, reducing data redundancy and improving the compactness and representativeness of the data.

[0034] It should be explained that the mode is the value with the highest frequency of occurrence and can reflect the user's preference setting at a specific indoor temperature. If there is a mode, it means that the user has a clear preference setting at this indoor temperature. Therefore, it is reasonable to choose the mode as the tendency set temperature. Of course, if there are multiple modes, take the tendency set temperature corresponding to the largest mode. When there is no mode, calculate the range of several set temperatures corresponding to the same indoor temperature. The range is the difference between the maximum value and the minimum value, which reflects the degree of dispersion of the data. By comparing the range with the allowable range, it can be judged whether the fluctuation range of the data is within the acceptable range. When the range is less than or equal to the allowable range, it means that the data fluctuation is small and the set temperatures of the user at this indoor temperature are relatively consistent. At this time, it is reasonable to take the average value as the tendency set temperature because the average value can represent the user's setting preference. When the range is greater than the allowable range, it means that the data fluctuation is large and the set temperatures of the user at this indoor temperature vary greatly. At this time, taking the maximum value as the tendency set temperature may increase energy consumption, but it is necessary to ensure user comfort in extreme cases.

[0035] Construct a coordinate system with the indoor temperature as the horizontal axis and the tendency set temperature as the vertical axis. Then, mark several points in the constructed coordinate system for the tendency set temperature corresponding to the indoor temperature to form a tendency correlation curve between the indoor temperature and the set temperature.

[0036] The set time tendency correlation module is used to classify the retrieved temperature setting records according to the same set time to generate a set time classification set, and based on this, analyze the preference set time and the tendency correlation between the preference set time and the set duration.

[0037] It should be emphasized that the set time in the temperature setting record reflects the heating demand time of users in the building. These demand times may have certain periodic preferences, that is, the heating demands of users are more concentrated within a specific time period. For example, users work outside during the day and prefer to enter the building at 6 pm after work. Therefore, 6 pm can be regarded as the preferred set time of users. Thus, load forecasting at this time point can more accurately reflect the true heating demands of users and improve the accuracy and reliability of forecasting. In addition, if load forecasting is performed for each set time in the temperature setting record, on the one hand, it will lead to complex data and reduce the forecasting efficiency; on the other hand, when the occurrence frequency of a certain set time is extremely low, performing load forecasting on it may result in invalid forecasting. Therefore, it is necessary to perform preference screening on the set time, and use the selected preferred set times as the key forecasting objects to improve the pertinence and accuracy of forecasting.

[0038] Preferably, the process of analyzing the preferred temperature set time is as follows: summarize the occurrence frequencies of the same set time, and then identify the distribution uniformity of the occurrence frequencies of each set time. If the identified distribution is not uniform, take the set time corresponding to the highest occurrence frequency as the preferred set time; otherwise, regard all set times as preferred set times.

[0039] In the operation of the above preferred implementation, the distribution uniformity identification can adopt the chi-square test method. Specifically, first make the null hypothesis and the alternative hypothesis, then calculate the total occurrence frequency of the set time, divide it by the number of set times to obtain the expected frequency, and then use the expected frequency to calculate the chi-square statistic. The calculation formula of the chi-square statistic is In the formula, O i represents the occurrence frequency of the i-th set time, i represents the set time number, E represents the expected frequency. Exemplarily, the null hypothesis is that the occurrence frequencies of each set time are evenly distributed, and the alternative hypothesis is that the occurrence frequencies of each set time are not evenly distributed. At this time, the set times retrieved from the temperature setting record are 18:00, 19:00, 18:30, 20:00, 18:00, 18:00, 19:00, 18:00, 18:30, 19:00, 18:00, 18:00. Among them, 18:00 appears 6 times, 19:00 appears 3 times, 18:30 appears 2 times, and 20:00 appears 1 time. The total occurrence frequency of the set time is 12 times, and the number of set times is 4. In this example, the expected frequency is The chi-square statistic is Furthermore, compare the calculated chi-square statistic with the critical value. If the chi-square statistic is greater than the critical value, reject the null hypothesis and consider that the frequency distribution is not uniform. If the chi-square statistic is less than or equal to the critical value, do not reject the null hypothesis and consider that the frequency distribution is uniform. The critical value can be obtained by looking up the table. For example, the critical value with a degree of freedom of 23 and a significance level of 0.05 is 35.17.

[0040] The present invention uses the chi-square test to identify whether the occurrence frequencies of each set time are uniformly distributed, which can provide an index of statistical significance and quantify the distribution characteristics of the data. This method is applicable not only to various types of categorical data, but also the results are intuitive and easy to understand, which helps to simplify data processing.

[0041] It should be understood that when identifying non-uniform distribution, it indicates that the occurrence frequencies of each set time vary greatly. At this time, the set time with the highest occurrence frequency reflects the set time selected by the user in most cases, which is usually the set time that the user is most accustomed to. Therefore, selecting the set time with the highest frequency can ensure that the set time has high representativeness. When identifying uniform distribution, it indicates that the occurrence frequencies of each set time are relatively close, and there is no single representative set time. At this time, all set times are used as the preferred set times.

[0042] Further preferably, the tendency association between the preferred set time and the set duration is as follows: Select the preferred set time classification set from the set time classification set, and extract the set duration from each temperature setting record included in the preferred set time classification set.

[0043] Similarly, the set duration of each temperature setting record in the preferred set time classification set is obtained according to the analysis method of the corresponding tendency set temperature for the indoor temperature to obtain the corresponding tendency set duration for the preferred set time.

[0044] It should be understood that the set time in the temperature setting record reflects the heating demand time of the user in the building, and the set duration corresponding to the set time reflects the heating demand duration of the user during the heating demand time. For example, the user prefers to enter the building at 6 pm and leave the building at 7 am, so the duration between 6 pm and 7 am is the heating demand duration of the user. By analyzing the tendency set duration corresponding to the preferred set time of the user, the heating demand duration of the user at different heating demand times can be reflected. The heating demand duration determines the operation duration of the heat exchange station. The longer the operation duration of the heat exchange station, the greater the impact on the load.

[0045] The heating load information retrieval module is used to retrieve the heating operation record related to the temperature setting record from the heating control center of the heat exchange station and extract the unit heating load therefrom.

[0046] It should be noted that the heating control center of the heat exchange station is the central control unit responsible for managing and controlling the entire heating system. It can record the heating operation status, including the operation duration, heating load, etc., during the heating operation when the user sets the temperature, and generate the heating operation record.

[0047] In a preferred implementation of the above solution, the following operations are involved in retrieving the heating operation records related to the temperature setting records: Extract the operation time and operation duration from the heating operation records retrieved from the heating control center of the heat exchange station, and match them with the set time and set duration in the temperature setting records. Thus, the successfully matched heating operation records are selected as the heating operation records related to the temperature setting records.

[0048] In the above, the operation time refers to the initial operation time. A successful match means that the operation time matches the set time and the operation duration matches the set duration. Here, a match means that the times are the same or close.

[0049] It should be noted that by using the operation time and operation duration in the heating operation records for matching, the heating operation records associated with the temperature setting records can be screened out. Since the temperature setting records reflect the heating demands of users, screening out the heating operation records associated with the temperature setting records can accurately reflect the operation conditions of the heat exchanger under each heating demand of users, which is beneficial to specifically extracting the heating load of the heat exchanger under each heating demand.

[0050] Furthermore, it should be noted that the unit heating load mentioned above refers to the heating load within the unit operation duration. Specifically, the heating load can be extracted from the heating operation records and then divided by the operation duration to obtain the unit heating load.

[0051] The heating temperature difference tendency correlation module is used to calculate the heating temperature difference by using the retrieved temperature setting records, and then analyze the tendency correlation between the heating temperature difference and the unit heating load.

[0052] In the way that the above solution can be implemented, the heating temperature difference is calculated as follows: Extract the set temperature from the temperature setting records, and subtract the synchronized indoor temperature mapped by the corresponding temperature setting records from it to obtain the heating temperature difference. Generally speaking, the greater the heating temperature difference, the greater the heating load.

[0053] Classify the heating temperature differences of the temperature setting records according to the same heating temperature difference to generate temperature setting record sets corresponding to each heating temperature difference.

[0054] Based on the heating operation records related to the temperature setting records, convert the temperature setting record sets corresponding to each heating temperature difference into heating operation record sets corresponding to each heating temperature difference.

[0055] For each heating operation record in the heating operation record sets corresponding to each heating temperature difference, in the same way as analyzing the tendency of the set temperature corresponding to the indoor temperature, obtain the tendency unit heating load corresponding to each heating temperature difference.

[0056] Taking the heating temperature difference as the horizontal axis and the inclined unit heating load as the vertical axis, a coordinate system is constructed. Thus, several points corresponding to the inclined unit heating load for each heating temperature difference are marked within the constructed coordinate system to form an inclined correlation curve between the heating temperature difference and the unit heating load.

[0057] The heating load prediction module is used to predict the heating load by comprehensively considering the inclined correlation between the indoor temperature and the set temperature, the preference setting time, the inclined correlation between the preference setting time and the set duration, and the inclined correlation between the heating temperature difference and the unit heating load.

[0058] In a further implementation that can be achieved by the above solution, the heating load prediction is carried out as follows: The building temperature control center is used to collect the indoor temperature in real time, and at the same time, the building human body sensing terminal is used to conduct human body sensing in real time. The human body sensing time is compared with the preference setting time. If the human body sensing time conforms to the preference setting time, the corresponding inclined set duration of the preference setting time is retrieved as the current inclined heating duration.

[0059] It should be noted that the building human body sensing terminal can adopt an infrared sensor for detecting human activities. When the infrared sensor senses a human body, it can be determined that the user has entered the building.

[0060] Capture the set temperature corresponding to the indoor temperature at the human body sensing time in the inclined correlation curve between the indoor temperature and the set temperature as the current inclined set temperature;

[0061] Match the current room temperature with the synchronized indoor temperature mapped by the temperature setting record. At the same time, match the current inclined set temperature, the current inclined heating duration, and the human body sensing time with the set temperature, the set time, and the set duration in the temperature setting record. If the matching is successful, the heating load involved in the heating operation record corresponding to the successfully matched temperature setting record is used as the predicted current heating load.

[0062] The above-mentioned successful matching means that the room temperature, the set temperature, the set time, and the set duration are all successfully matched. The above-mentioned heating load refers to the total heating load.

[0063] It should be explained that when the current room temperature, the current inclined set temperature, the current inclined heating duration, and the human body sensing time are all matched with a certain temperature setting record, it indicates that the current user's heating demand is highly consistent with a certain historical heating demand. In this case, directly predicting the heating load in the heating operation record involved in the matched temperature setting record can improve the prediction accuracy, reduce the calculation complexity, and improve the system response speed.

[0064] When the matching fails, it indicates that there is no situation where the current user's heating demand is highly consistent with the historical heating demands. At this time, capture the set temperature corresponding to the indoor temperature at the human induction time in the tendency correlation curve between the indoor temperature and the set temperature, and calculate the difference between this set temperature and the indoor temperature corresponding to the human induction time to obtain the current tendency heating temperature difference.

[0065] Capture the tendency unit heating load corresponding to the current heating temperature difference from the tendency correlation curve between the heating temperature difference and the unit heating load.

[0066] Multiply the current tendency heating duration by the tendency unit heating load corresponding to the current heating tendency temperature difference to predict the current heating load.

[0067] It further needs to be explained that when there is no situation where the current user's heating demand is highly consistent with the historical heating demands, at this time, focus on the heating temperature difference and the heating duration. The heating temperature difference reflects the gap between the current indoor temperature and the desired temperature, and is an important factor affecting the heating load. The heating duration reflects the actual demand time of the user in the building and directly affects the operating load of the heating system. Combining the temperature difference and the duration can more comprehensively evaluate the heating demand. Even if the current user's heating demand is not completely consistent with the historical data, by paying attention to the temperature difference and the duration, the accuracy of load prediction can be maximally improved, and at the same time, complex mathematical models and algorithms are avoided, simplifying the process of load prediction.

[0068] In the innovative implementation of the above solution, compare the human induction time with the preference setting time. If the human induction time does not conform to the preference setting time, it indicates that the time when the current user enters the building does not belong to the regular tendency time. In this case, the human induction time may be higher or lower than the preference setting time. At this time, take the average value of the tendency setting durations corresponding to each preference setting time to obtain the predicted current tendency heating duration, which can provide a relatively stable reference value and avoid prediction deviations caused by individual abnormal data.

[0069] In the above situation, there is a situation where the current user's heating demand is consistent with the historical heating demand.

[0070] Combine the predicted current tendency heating duration with the tendency unit heating load corresponding to the current heating tendency temperature difference to predict the current heating load. Specifically, refer to the method of matching failure for heating load prediction in the same way.

[0071] The above content is only an example and explanation of the concept of the present invention. Those skilled in the art of this technology make various modifications or supplements to the described specific embodiments or use similar methods for substitution. As long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they should fall within the protection scope of the present invention.

Claims

1. A heat exchange station load prediction and analysis system based on deep learning, characterized in that ,include: The indoor temperature acquisition module extracts the real-time indoor temperature collected from the building temperature control center during the heating period to generate an indoor temperature time series; The temperature setting information extraction module retrieves the user's temperature setting records from the building temperature control center during the heating period, and extracts the setting time, setting duration and setting temperature from them; A synchronous indoor temperature extraction module extracts the indoor temperature corresponding to the set time from the indoor temperature time series based on the set time as the synchronous indoor temperature; The set temperature tendency association module maps the user's temperature setting records with the synchronized indoor temperature, classifies the temperature setting records corresponding to the same synchronized indoor temperature to generate an indoor temperature classification set, and then analyzes the tendency association between the indoor temperature and the set temperature based on this; The setting time tendency association module classifies the retrieved temperature setting records according to the same setting time to generate a setting time classification set, and analyzes the preferred setting time and the tendency association between the preferred setting time and the setting duration; The heating load information retrieval module retrieves the heating operation records involved in the temperature setting records from the heating control center of the heat exchange station, and extracts the unit heating load from them; The heating temperature difference tendency correlation module calculates the heating temperature difference using the retrieved temperature setting records, and then analyzes the tendency correlation between the heating temperature difference and the unit heating load; The heating load prediction module predicts the heating load by comprehensively considering the tendency correlation between the indoor temperature and the set temperature, the preferred setting time and the tendency correlation between the preferred setting time and the set duration, and the tendency correlation between the heating temperature difference and the unit heating load; The analysis of the trend correlation between the indoor temperature and the set temperature is shown in the following process: Extracting the set temperature from each temperature setting record contained in the indoor temperature classification set; Compare several set temperatures to identify whether there is a mode. If there is a mode, the mode is used as the tendency set temperature. If there is no mode, the range of several set temperatures is calculated and compared with the allowable range of the system configuration. If the calculated range is less than or equal to the allowable range, the average of several set temperatures is taken as the tendency set temperature. Otherwise, the maximum value of several set temperatures is taken as the tendency set temperature. A coordinate system is constructed with the indoor temperature as the horizontal axis and the tendency to the set temperature as the vertical axis, and several points are marked in the coordinate system to form a tendency correlation curve between the indoor temperature and the set temperature; The tendency correlation between the preferred setting time and the setting duration is shown in the following analysis process: Selecting a preferred setting time classification set from the setting time classification set, and extracting the setting duration from each temperature setting record included in the preferred setting time classification set; Similarly, the setting time length is analyzed in the same way as the indoor temperature corresponding to the preferred setting temperature to obtain the preferred setting time length corresponding to the preferred setting time; The analysis of the heating temperature difference and the tendency of the unit heating load is as follows: Classify the heating temperature differences of the temperature setting records according to the same heating temperature difference to generate a temperature setting record set corresponding to each heating temperature difference; Based on the heating operation record involved in the temperature setting record, the temperature setting record set corresponding to each heating temperature difference is converted into a heating operation record set; The unit heating load of each heating operation record in the heating operation record set is similarly analyzed in the same way as the indoor temperature corresponding to the tendency set temperature to obtain the tendency unit heating load corresponding to each heating temperature difference; A coordinate system is constructed with the heating temperature difference as the horizontal axis and the tendency of the unit heating load as the vertical axis, and several points are marked in the constructed coordinate system to form a tendency correlation curve between the heating temperature difference and the unit heating load.

2. A heat exchange station load prediction and analysis system based on deep learning as claimed in claim 1, characterized in that: The analysis preference setting time is as follows: The occurrence frequencies of the same setting time are summarized, and then the occurrence frequencies of each setting time are identified for distribution uniformity. If the distribution is identified to be uneven, the setting time corresponding to the highest occurrence frequency is taken as the preferred setting time, otherwise all setting times are taken as the preferred setting time.

3. The heat exchange station load prediction and analysis system based on deep learning according to claim 1, characterized in that: The heating operation record involved in retrieving the temperature setting record is performed as follows: The operation time and operation duration are extracted from the heating operation records retrieved from the heating control center of the heat exchange station, and matched with the set time and set duration in the temperature setting record, thereby selecting the heating operation records with successful matching as the heating operation records involved in the temperature setting record.

4. The heat exchange station load prediction and analysis system based on deep learning according to claim 1, characterized in that: The calculation of the heating temperature difference using the retrieved temperature setting record is implemented as follows: The set temperature is extracted from the temperature setting record, and the heating temperature difference is obtained by subtracting the set temperature from the synchronous indoor temperature mapped by the corresponding temperature setting record.

5. The heat exchange station load prediction and analysis system based on deep learning according to claim 1, characterized in that: The heating load forecasting is implemented as follows: The indoor temperature is collected in real time by the building temperature control center, and the human body sensing terminal is used to sense the human body in real time, and the human body sensing time is compared with the preferred setting time. If the human body sensing time is consistent with the preferred setting time, the preferred setting time corresponding to the preferred setting time is retrieved as the current preferred heating time. The set temperature corresponding to the indoor temperature in the tendency correlation curve of the human body sensing time is captured from the tendency correlation curve of the indoor temperature and the set temperature as the current tendency set temperature; The current room temperature is matched with the synchronized indoor temperature mapped by the temperature setting record, and the current tendency setting temperature, the current tendency heating time and the human body sensing time are matched with the set temperature, the set time and the set time in the temperature setting record. If the match is successful, the heating load of the heating operation record involving the successfully matched temperature setting record is used as the predicted current heating load.

6. A heat exchange station load prediction and analysis system based on deep learning as claimed in claim 5, characterized in that: The heating load forecasting also includes the following process: When the matching fails, the set temperature corresponding to the indoor temperature at the time of human body induction is captured from the trend correlation curve of the indoor temperature and the set temperature, and the difference is made with the indoor temperature corresponding to the time of human body induction to obtain the current trend heating temperature difference; The tendency unit heating load corresponding to the current heating temperature difference is captured from the tendency correlation curve between the heating temperature difference and the unit heating load; The current heating load is predicted by combining the current heating load duration with the unit heating load corresponding to the current heating temperature difference.

7. A heat exchange station load prediction and analysis system based on deep learning as claimed in claim 6, characterized in that: The heating load forecasting further includes the following process: Compare the human body sensing time with the preference setting time. If the human body sensing time does not match the preference setting time, then average the preference setting time corresponding to each preference setting time to obtain the predicted current preference heating time. The current heating load is predicted by combining the predicted current tendency heating time with the tendency unit heating load corresponding to the current tendency heating temperature difference.

Images