A Method for Predicting the Daily Load of Fixed-Frequency Central Air Conditioners Considering the Operating Status of Air Conditioners

Abstract

The present invention provides a day-ahead load forecasting method for a fixed-frequency central air conditioner considering the operating state of the air conditioner. The method includes obtaining n sets of monitoring data in a historical period; performing normalization processing on the monitoring data to obtain dimensionless new data, and respectively performing correlation analysis between the new data of the factors affecting the cooling capacity demand of the central air conditioner and the load rate of the central air conditioner system, and allocating weights according to the results of the correlation analysis to obtain a historical data sequence of contribution values; establishing an EPGA-BPNN model; obtaining a data sequence of contribution values in a set period before the forecasting period; inputting the trained EPGA-BPNN model to obtain the load rate of the central air conditioner system in the forecasting period; determining the operating state of the central air conditioner system in the forecasting period, and formulating a corresponding regulation strategy for the forecasting period. This method can improve the accuracy of the day-ahead load forecasting of the central air conditioner, can achieve precise regulation of the central air conditioner, and is applicable to the forecasting of air conditioner load and operation stage.

Description

Technical Field

[0001] The present invention relates to a method for predicting the daily load of a fixed-frequency central air conditioner considering the operating state of the air conditioner, belonging to the technical field of central air conditioner load prediction. Background Technique

[0002] With the rapid development of the Chinese economy, the building scale in cities is continuously expanding. As one of the main energy-consuming equipment in buildings, the electricity consumption of central air conditioning systems accounts for a large proportion of urban electricity consumption. In response to the "dual carbon" goal, research on central air conditioner load prediction and control strategies can greatly reduce its own energy consumption and well achieve the purpose of reducing air conditioner operating costs, energy conservation and emission reduction.

[0003] Central air conditioner load prediction generally uses mathematical models and algorithms to predict the load of the central air conditioning system in a future period. Central air conditioner load prediction usually involves multiple factors affecting the air conditioner state, including indoor and outdoor temperature and humidity, wind speed, human flow, building itself conditions, etc. By collecting and analyzing these data, a prediction model can be established. The methods for establishing a prediction model include mechanism modeling, data-driven modeling, etc. Mechanism modeling is also called white-box modeling, which is to establish a model according to the internal structure principle of the object under study, the changes of each heat transfer medium, and each external related factor. The model established based on the mechanism is very cumbersome, and a large number of differential equations need to be calculated when designing the controller, so there will be great limitations in engineering applications. Data-driven modeling is also called black-box modeling. The data-driven model uses system experimental data for modeling, does not require a lot of theoretical knowledge, and only needs to analyze the experimental data and establish the model structure through the relationship between input and output.

[0004] At present, the research on central air conditioner load prediction models and control strategies is still in the preliminary stage. The accuracy of central air conditioner load resource collection is not high, and the refined control level is insufficient. There is an urgent need to explore more accurate prediction methods for central air conditioner loads. Therefore, researching a method for predicting the daily load of a fixed-frequency central air conditioner considering the operating state of the air conditioner is beneficial to the accurate control of central air conditioner loads and realizes flexible interaction with power grid dispatching.

[0005] The above problems should be considered and solved in the process of predicting the daily load of a fixed-frequency central air conditioner considering the operating state of the air conditioner. Summary of the Invention

[0006] The purpose of the present invention is to provide a method for predicting the daily load of a fixed-frequency central air conditioner considering the operating state of the air conditioner to solve the problem that the accuracy of the existing central air conditioner daily load prediction needs to be improved.

[0007] The technical solution of the present invention is:

[0008] A method for predicting the daily load of a fixed-frequency central air conditioner considering the operating state of the air conditioner, comprising the following steps:

[0009] S1. Obtain n groups of monitoring data in a historical period. Each group of monitoring data includes factors affecting the cooling capacity demand of the central air conditioner and the total operating power of the central air conditioner system;

[0010] S2. Perform normalization processing on the monitoring data to obtain dimensionless new data, and respectively perform correlation analysis between the new data of the factors affecting the cooling capacity demand of the central air conditioner and the load rate of the central air conditioner system. According to the results of the correlation analysis, assign weights to obtain the historical data sequence of the contribution values of each dimensionless new data to the load rate of the central air conditioner system;

[0011] S3. Establish an improved backpropagation neural network model based on the elite retention strategy genetic algorithm EPGA, namely the EPGA-BPNN model. Use the historical data sequence of the contribution values as input data, and combine the load rate of the central air conditioner system in the historical period. After training the EPGA-BPNN model, obtain the trained EPGA-BPNN model;

[0012] S4. Obtain the monitoring data in a set period before the prediction period, and obtain the data sequence of the contribution values in the set period before the prediction period;

[0013] S5. Input the data sequence of the contribution values in the set period before the prediction period into the trained EPGA-BPNN model to obtain the load rate of the central air conditioner system in the prediction period;

[0014] S6. According to the load rate of the central air conditioner system in the prediction period obtained in step S5, determine the operating state of the central air conditioner system in the prediction period, and formulate corresponding regulation strategies for the prediction period according to the operating state.

[0015] Further, in step S1, when obtaining n groups of monitoring data in a historical period, each group of monitoring data includes factors affecting the cooling capacity demand of the central air conditioner and the total operating power of the central air conditioner system. Specifically, collect n groups of factors affecting the cooling capacity demand of the central air conditioner, including indoor temperature T in , indoor humidity H in , outdoor temperature T out , outdoor humidity H out and chilled water temperature T cw ; and collect the corresponding total operating power of the central air conditioner system as P air ; obtain n groups of monitoring data .

[0016] Further, in step S2, when performing normalization processing on the monitoring data to obtain dimensionless new data, specifically, perform normalization processing on the monitoring data:

[0017] ,

[0018] Among them, the sequences respectively represent the input variables indoor temperature Tin, indoor humidity Hin, outdoor temperature Tout, outdoor humidity Hout, and chilled water temperature Tcw. The sequence is the data sequence of the total operating power Pair of the central air-conditioning system, that is , and after transformation, the output quantity is obtained. This output quantity is a dimensionless new data.

[0019] Furthermore, in step S2, the new data of the factors affecting the cooling capacity demand of the central air-conditioning system are respectively analyzed for their correlation with the load rate of the central air-conditioning system:

[0020] ,

[0021] Among them, , is the load rate sequence of the central air-conditioning system and is the mother sequence; is the i-th standardized output sequence and is the child sequence; represents the correlation between the i-th child sequence and the corresponding dimension of the mother sequence, and its value range is between 0 and 1. 0 means no correlation, 1 means strong correlation, and the larger the number, the stronger the correlation; is the resolution coefficient; p and q are the minimum difference and the maximum difference between two levels respectively: , .

[0022] Furthermore, in step S2, according to the correlation analysis results, weights are assigned to obtain the historical data sequence of the contribution values of each dimensionless new data to the load rate of the central air-conditioning system. Specifically, according to the correlation analysis results, weights [[ID=3⑨]]are assigned, and the normalized data sequence is multiplied by the assigned weight value to obtain the historical data sequence of the contribution values to the load rate: .

[0023] Furthermore, in step S3, the backpropagation neural network model improved by the elite retention strategy genetic algorithm EPGA, that is, the EPGA-BPNN model, includes the backpropagation neural network model, that is, the BPNN model and the parameter optimization module.

[0024] BPNN model: The historical data sequence of the contribution values from the w-v moment to the w moment in the historical period is used as the input data for forward propagation, and the predicted load rate from the w + 1 moment to the w + u moment is obtained; the error between the obtained predicted load rate and the corresponding total operating power of the central air-conditioning system in the historical period is propagated backward.

[0025] Parameter optimization module: The elitist retention strategy genetic algorithm (EPGA) is used to iteratively optimize the network parameters of the BPNN model to obtain the optimized network parameters of the BPNN model.

[0026] Further, in step S3, using the contribution value historical data sequence as input data and combining with the central air-conditioning system load rate in the historical period, the EPGA-BPNN model is trained to obtain the trained EPGA-BPNN model. Specifically,

[0027] S31: Input the contribution value historical data sequence from time w - v to time w in the historical period into the BPNN model of the EPGA-BPNN model, and output the predicted value sequence of the central air-conditioning system load rate from time w + 1 to time w + u;

[0028] S32: Calculate the loss function Loss of the prediction error from the predicted value sequence of the central air-conditioning system load rate and the total operating power of the central air-conditioning system in the historical period;

[0029] S33: Initialize the network parameters of the BPNN model, and use the initialized network parameters as the initial parameter population;

[0030] S34: When the loss function Loss of the prediction error is backpropagated, use the elitist retention strategy genetic algorithm (EPGA) to iteratively update the parameter population. Repeat steps S31 - S34 until the loss function is less than the set value or reaches the set number of times, and then solve to obtain the optimal network parameter values;

[0031] S35: Assign the optimal network parameter values to the BPNN model to obtain the trained EPGA-BPNN model.

[0032] Further, in step S34, when the loss function Loss of the prediction error is backpropagated, use the elitist retention strategy genetic algorithm (EPGA) to iteratively update the parameter population. Specifically,

[0033] S341: Denote the parameter population of the elitist retention strategy genetic algorithm (EPGA) as , where x represents the population individual and t represents the population generation;

[0034] S342: In each iteration, use the loss function Loss as the fitness function value of the individual, and directly copy the best individual in the population, that is, the individual with the highest fitness function value, to the next generation;

[0035] S343: On the premise of keeping the overall scale of the old population unchanged, combine the new individuals generated by the genetic operator operation of the other individuals of the old population with the best individual in the population to form the new population ; For the new When evolving again, recalculate the fitness function value of the individual; use the best-performing individual in the population as the optimal network parameter value.

[0036] Furthermore, in step S6, the operating states of the central air-conditioning system during the prediction period include the startup stage, the running stage, and the stop stage.

[0037] Furthermore, in step S6, determine the operating state of the central air-conditioning system during the prediction period according to the load rate of the central air-conditioning system obtained in step S5. Specifically,

[0038] Among the load rates during the prediction period, any one of the following three situations is determined as the startup stage:

[0039] Situation 1: The load rate shows a continuous increase, and the rising change rate of the load rate at the starting moment and the ending moment is within the first set interval;

[0040] Situation 2: There is a period during the prediction period when the load rate reaches within the second set interval;

[0041] Situation 3: The load rate shows a continuous decrease, and the falling change rate of the load rate at the starting moment and the ending moment is within the third set interval;

[0042] When the load rate during the prediction period remains within the set range and the load rate change rate is within the fourth set interval, it is determined as the running stage;

[0043] The falling change rate of the load rate at the starting moment and the ending moment is within the fifth set interval, and the load rate drops to 0, which is determined as the stop stage.

[0044] The beneficial effects of the present invention are as follows: This kind of day-ahead load prediction method for fixed-frequency central air-conditioning considering the air-conditioning operating state aims at the day-ahead load prediction of fixed-frequency central air-conditioning, considers the air-conditioning operating state, uses the EPGA-BPNN model to predict the air-conditioning load rate, and judges the air-conditioning operating stage during the prediction period according to the load rate, which can improve the accuracy of the day-ahead load prediction of central air-conditioning, can achieve precise control of central air-conditioning, and is applicable to the prediction of air-conditioning load and operating stage. Brief Description of the Drawings

[0045] Figure 1 is a schematic flowchart of the day-ahead load prediction method for fixed-frequency central air-conditioning considering the air-conditioning operating state in an embodiment of the present invention;

[0046] Figure 2 is a schematic illustration of the iterative update of the parameter population using the elitist retention strategy genetic algorithm EPGA in the embodiment;

[0047] Figure 3It is a schematic diagram of the change trend of the load rate corresponding to the operating state of the central air-conditioning system in the embodiment. Detailed implementation manners

[0048] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0049] Embodiment

[0050] A day-ahead load forecasting method for a fixed-frequency central air conditioner considering the operating state of the air conditioner, as Figure 1 , includes the following steps.

[0051] S1. Obtain n sets of monitoring data in a historical period. Each set of monitoring data includes factors affecting the cooling capacity demand of the central air-conditioning system and the total operating power of the central air-conditioning system.

[0052] In step S1, to obtain n sets of monitoring data in a historical period, each set of monitoring data includes factors affecting the cooling capacity demand of the central air-conditioning system and the total operating power of the central air-conditioning system. Specifically, collect n sets of factors affecting the cooling capacity demand of the central air-conditioning system, including indoor temperature T in , indoor humidity H in , outdoor temperature T out , outdoor humidity H out and chilled water temperature T cw， , which are collected by temperature and humidity sensors; and collect the corresponding total operating power of the central air-conditioning system as P air ; to obtain n sets of monitoring data .

[0053] S2. Normalize the monitoring data to obtain dimensionless new data, and respectively perform correlation analysis between the new data of the factors affecting the cooling capacity demand of the central air-conditioning system and the load rate of the central air-conditioning system. According to the results of the correlation analysis, assign weights to obtain the historical data sequence of the contribution values of each dimensionless new data to the load rate of the central air-conditioning system.

[0054] In step S2, to normalize the monitoring data to obtain dimensionless new data, specifically, normalize the monitoring data:

[0055] ,

[0056] where the sequences respectively represent the input variables indoor temperature Tin, indoor humidity Hin, outdoor temperature Tout, outdoor humidity Hout, chilled water temperature Tcw, and the sequence is the data sequence of the total operating power of the central air-conditioning system as Pair, that is , and after transformation, the output quantity is obtained, and this output quantity is dimensionless new data.

[0057] In step S2, new data on factors affecting the cooling capacity demand of the central air conditioner are respectively subjected to correlation analysis with the load rate of the central air conditioner system:

[0058] ,

[0059] Among them, , is the load rate sequence of the central air conditioner system and is the mother sequence; is the i-th normalized output sequence and is the child sequence; represents the correlation between the i-th child sequence and the corresponding dimension of the mother sequence, and its value range is between 0 and 1. 0 indicates no correlation, 1 indicates strong correlation, and the larger the number, the stronger the correlation; is the resolution coefficient; p and q are the two-level minimum difference and the two-level maximum difference respectively: , .

[0060] In step S2, weights are assigned according to the correlation analysis results to obtain a historical data sequence of the contribution values of each dimensionless new data to the load rate of the central air conditioner system. Specifically, weights are assigned according to the correlation analysis results , and the normalized data sequence is multiplied by the assigned weight value to obtain a historical data sequence of the contribution values to the load rate: .

[0061] S3. Establish an improved backpropagation neural network model based on the elitist retention strategy genetic algorithm EPGA, namely the EPGA-BPNN model. Use the historical data sequence of the contribution values as input data, and combine the load rate of the central air conditioner system in the historical period to train the EPGA-BPNN model to obtain the trained EPGA-BPNN model.

[0062] In step S3, the improved backpropagation neural network model based on the elitist retention strategy genetic algorithm EPGA, namely the EPGA-BPNN model, includes a backpropagation neural network model, namely the BPNN model, and a parameter optimization module.

[0063] BPNN model: Use the historical data sequence of the contribution values from the w-v moment to the w moment in the historical period as input data, perform forward propagation, and obtain the predicted load rate from the w+1 moment to the w+u moment; Backpropagate the error between the obtained predicted load rate and the corresponding total operating power of the central air conditioner system in the historical period.

[0064] Parameter optimization module: Use the elitist retention strategy genetic algorithm EPGA to iteratively optimize the network parameters of the BPNN model to obtain the network parameters of the optimized BPNN model.

[0065] In step S3, using the historical contribution value data sequence as input data and combining with the load rate of the central air-conditioning system in the historical period, the EPGA-BPNN model is trained to obtain the trained EPGA-BPNN model. Specifically,

[0066] S31. Input the historical contribution value data sequence from time w-v to time w in the historical period into the BPNN model of the EPGA-BPNN model, and output the predicted value sequence of the load rate of the central air-conditioning system from time w+1 to time w+u;

[0067] S32. Calculate the loss function Loss of the prediction error from the predicted value sequence of the load rate of the central air-conditioning system and the total operating power of the central air-conditioning system in the historical period;

[0068] S33. Initialize the network parameters of the BPNN model, and use the initialized network parameters as the initial parameter population;

[0069] S34. When the loss function Loss of the prediction error is backpropagated, use the elitist-preserving genetic algorithm EPGA to iteratively update the parameter population. Repeat steps S31-S34 until the loss function is less than the set value or reaches the set number of times, and then solve to obtain the optimal network parameter values;

[0070] In step S34, when the loss function Loss of the prediction error is backpropagated, use the elitist-preserving genetic algorithm EPGA to iteratively update the parameter population. Specifically, Figure 2 :

[0071] S341. Denote the parameter population of the elitist-preserving genetic algorithm EPGA as , where x represents the population individual and t represents the population generation;

[0072] S342. In each iteration, use the loss function Loss as the fitness function value of the individual, and directly copy the best individual in the population, that is, the individual with the highest fitness function value, to the next generation;

[0073] S343. On the premise of keeping the overall scale of the old population unchanged, form a new population by combining the new individuals generated by the genetic operator operation of the other individuals in the old population and the best individual in the population; recalculate the fitness function value of the individual when evolving the new again; use the best individual in the population as the optimal network parameter value.

[0074] S35. Assign the optimal network parameter values to the BPNN model to obtain the trained EPGA-BPNN model.

[0075] In step S3, the parameter optimization module uses the elitist retention strategy genetic algorithm (EPGA) to solve the optimal network parameters of the BPNN model. Each individual in the EPGA algorithm population represents a set of network parameters of the BPNN model, including all the weights and thresholds of the BPNN model. When the elitist retention strategy genetic algorithm (EPGA) iterates, the mean square error function Loss between the BPNN output value and the true value is used as the fitness function to evaluate the performance of individuals.

[0076] In step S3, the population of the elitist retention strategy genetic algorithm (EPGA) is denoted as (where x represents the population individual and t represents the population generation). EPGA uses an iterative update method to find the optimal individual. In each iteration, the best-performing individual in the population is directly copied to the next generation; the old population while keeping the overall scale unchanged, retains the best-performing set of network parameters in according to the calculation result of the Loss function, and together with the new individuals generated by other individuals through genetic operator operations, forms a new population , and recalculates the fitness function value of the individual when evolving again.

[0077] S4. Obtain the monitoring data for a set period before the prediction period, and obtain the contribution value data sequence for the set period before the prediction period.

[0078] In step S4, for the monitoring data obtained for the set period before the prediction period, the contribution value data sequence for the set period before the prediction period is obtained by using the same method as in step S2.

[0079] S5. Input the contribution value data sequence for the set period before the prediction period into the trained EPGA-BPNN model to obtain the load rate of the central air-conditioning system for the prediction period.

[0080] In step S5, the contribution value data sequence from T-n to T is input into the trained EPGA-BPNN model to predict the air-conditioning load rate for the prediction period, i.e., from T + 1 to T + m .

[0081] In step S5, the load rate of the central air-conditioning system = the load of the central air-conditioning system / the total rated power of each device in the central air-conditioning system. The total rated power of each device in the central air-conditioning system is a fixed value, so predicting the load rate of the central air-conditioning system is also called central air-conditioning load prediction.

[0082] S6. According to the load rate of the central air-conditioning system for the prediction period obtained in step S5, determine the operating state of the central air-conditioning system for the prediction period, and formulate corresponding regulation strategies for the prediction period according to the operating state.

[0083] In step S6, the operating states of the central air-conditioning system during the prediction period include the startup stage, the running stage, and the stop stage.

[0084] In step S6, based on the load rate of the central air-conditioning system during the prediction period obtained in step S5, the operating state of the central air-conditioning system during the prediction period is determined. Specifically,

[0085] Among the load rates during the prediction period, any one of the following three cases is determined as the startup stage:

[0086] Case 1: The load rate shows a continuous increase, and the rising change rate of the load rate at the starting moment and the ending moment is within the first set interval, such as , with the interval unit being % / 1h;

[0087] Case 2: There is a period during the prediction period when the load rate reaches the second set interval, such as within, with the interval unit being %;

[0088] Case 3: The load rate shows a continuous decrease, and the falling change rate of the load rate at the starting moment and the ending moment is within the third set interval, such as within, with the interval unit being %;

[0089] When the load rate during the prediction period remains within the set range, such as 50% to 80%, and the load rate change rate is within the fourth set interval, such as within, with the interval unit being % / 1h, it is determined as the running stage;

[0090] The falling change rate of the load rate at the starting moment and the ending moment is within the fifth set interval within, with the interval unit being % / 1h, and the load rate drops to 0, it is determined as the stop stage.

[0091] This method for predicting the daily load of a fixed-frequency central air-conditioning system considering the air-conditioning operating state aims at predicting the daily load of a fixed-frequency central air-conditioning system. Considering the air-conditioning operating state, it uses the EPGA-BPNN model to predict the air-conditioning load rate, and judges the air-conditioning operating stage during the prediction period according to the load rate, which can improve the accuracy of the daily load prediction of the central air-conditioning system, can achieve precise control of the central air-conditioning system, and is applicable to the prediction of air-conditioning load and operating stage.

[0092] The above are only the preferred embodiments of the present invention, but do not limit the patent scope of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions described in the foregoing specific embodiments, or perform equivalent replacements on some of the technical features. Any equivalent structures made by using the content of the specification and drawings of the present invention, directly or indirectly applied in other related technical fields, are equally within the scope of the patent protection of the present invention.

Claims

1. A day-ahead load forecasting method for fixed-frequency central air conditioners considering the operating status of air conditioners, characterized in that: including the following steps, S1. Obtain n groups of monitoring data in a historical period. Each group of monitoring data includes factors affecting the cooling capacity demand of the central air-conditioning system and the total operating power of the central air-conditioning system; S2. Perform normalization processing on the monitoring data to obtain dimensionless new data, and respectively perform correlation analysis between the new data of the factors affecting the cooling capacity demand of the central air-conditioning system and the load rate of the central air-conditioning system. According to the results of the correlation analysis, assign weights to obtain the historical data sequence of the contribution values of each dimensionless new data to the load rate of the central air-conditioning system; S3. Establish a backpropagation neural network model improved by the elitist retention strategy genetic algorithm EPGA, that is, the EPGA-BPNN model. Use the historical data sequence of contribution values as input data, and combine the load rate of the central air-conditioning system in the historical period to train the EPGA-BPNN model to obtain the trained EPGA-BPNN model; In step S3, the backpropagation neural network model improved by the elitist retention strategy genetic algorithm EPGA, that is, the EPGA-BPNN model, includes a backpropagation neural network model, that is, the BPNN model and a parameter optimization module; BPNN model: Use the historical data sequence of contribution values from the w-v moment to the w moment in the historical period as input data, perform forward propagation, and obtain the predicted load rate from the w+1 moment to the w+u moment; Backpropagate the error between the obtained predicted load rate and the corresponding total operating power of the central air-conditioning system in the historical period; Parameter optimization module: Use the elitist retention strategy genetic algorithm EPGA to iteratively optimize the network parameters of the BPNN model to obtain the network parameters of the optimized BPNN model; S4. Obtain the monitoring data in a set period before the prediction period and obtain the data sequence of contribution values in a set period before the prediction period; S5. Input the data sequence of contribution values in a set period before the prediction period into the trained EPGA-BPNN model to obtain the load rate of the central air-conditioning system in the prediction period; S6. According to the load rate of the central air-conditioning system in the prediction period obtained in step S5, determine the operating state of the central air-conditioning system in the prediction period, and formulate corresponding regulation strategies for the prediction period according to the operating state.

2. The day-ahead load forecasting method for fixed-frequency central air conditioners considering the operating status of air conditioners according to claim 1, wherein: In step S1, n sets of monitoring data for a historical period are obtained. Each set of monitoring data includes factors affecting the cooling capacity demand of the central air conditioner and the total operating power of the central air conditioning system. Specifically, the factors collected for the n sets of cooling capacity demand of the central air conditioner include indoor temperature T in , indoor humidity H in , outdoor temperature T out , outdoor humidity H out , and chilled water temperature T cw ; and the total operating power of the corresponding central air conditioning system is collected as P air ; obtaining n sets of monitoring data .

3. The day-ahead load forecasting method for fixed-frequency central air conditioners considering the operating status of air conditioners according to claim 1, characterized in that: In step S2, perform normalization processing on the monitoring data to obtain dimensionless new data. Specifically, normalize the monitoring data: ， Among them, , the sequence respectively represents the input variables indoor temperature Tin, indoor humidity Hin, outdoor temperature Tout, outdoor humidity Hout, chilled water temperature Tcw, and the sequence is the data sequence of the total operating power Pair of the central air-conditioning system, that is , after transformation, the output quantity is obtained, and this output quantity is a dimensionless new data.

4. The day-ahead load forecasting method for a fixed-frequency central air conditioner considering the operating state of the air conditioner according to any one of claims 1-3, characterized in that: In step S2, respectively perform correlation analysis between the new data of the factors affecting the cooling capacity demand of the central air-conditioning system and the load rate of the central air-conditioning system: ， Among them, , is the load rate sequence of the central air-conditioning system, which is the mother sequence; is the i-th normalized output sequence, which is the sub-sequence; represents the correlation between the i-th sub-sequence and the corresponding dimension of the mother sequence. The value range is between 0 and 1. 0 means no correlation, 1 means strong correlation. The larger the number, the stronger the correlation; is the resolution coefficient; p and q are the minimum difference and the maximum difference of two levels respectively: , .

5. The day-ahead load forecasting method for a fixed-frequency central air conditioner considering the operating status of the air conditioner according to any one of claims 1-3, characterized in that: In step S2, weights are assigned according to the correlation analysis results to obtain the historical data sequence of the contribution values of each dimensionless new data to the load rate of the central air-conditioning system. Specifically, weights are assigned according to the correlation analysis results , and the data sequence after normalization is multiplied by the assigned weight value to obtain the historical data sequence of the contribution values to the load rate: 。 6. The day-ahead load forecasting method for a fixed-frequency central air conditioner considering the operating state of the air conditioner according to any one of claims 1-3, characterized in that: In step S3, use the historical data sequence of contribution values as input data, and combine the load rate of the central air-conditioning system in the historical period to train the EPGA-BPNN model to obtain the trained EPGA-BPNN model. Specifically, S31. Input the historical data sequence of contribution values from the w-v moment to the w moment in the historical period into the BPNN model of the EPGA-BPNN model, and output the predicted value sequence of the load rate of the central air-conditioning system from the w+1 moment to the w+u moment; S32. Calculate the loss function Loss of the prediction error from the predicted value sequence of the load rate of the central air-conditioning system and the total operating power of the central air-conditioning system in the historical period; S33. Initialize the network parameters of the BPNN model, and use the initialized network parameters as the initial parameter population; S34. When the loss function Loss of the prediction error is backpropagated, use the elitist retention strategy genetic algorithm EPGA to iteratively update the parameter population. Repeat steps S31 - S34 until the loss function is less than the set value or the set number of times is reached, and then solve to obtain the optimal network parameter value; S35. Assign the optimal network parameter value to the BPNN model to obtain the trained EPGA - BPNN model.

7. The day-ahead load forecasting method for fixed-frequency central air conditioners considering the operating status of air conditioners according to claim 6, characterized in that: In step S34, when the loss function Loss of the prediction error is backpropagated, use the elitist retention strategy genetic algorithm EPGA to iteratively update the parameter population. Specifically, S341. The parameter population of the elitist retention strategy genetic algorithm EPGA is denoted as , where x represents an individual in the population and t represents the generation number of the population. S342. In each iteration, use the loss function Loss as the fitness function value of the individual, and directly copy the best - performing individual in the population, that is, the individual with the highest fitness function value, to the next generation; S343, old population Under the premise of keeping the overall size unchanged, the old population The new individuals generated by the genetic operator and the best performing individuals in the population together form a new population. ; for new When evolving again, the fitness function value of the individual is recalculated; the best performing individual in the population is taken as the optimal network parameter value.

8. The day-ahead load forecasting method for the fixed-frequency central air conditioner considering the operating status of the air conditioner according to any one of claims 1-3, characterized in that: In step S6, the operating states of the central air - conditioning system during the prediction period include the startup stage, the running stage, and the stop stage.

9. The day-ahead load forecasting method for fixed-frequency central air conditioners considering the operating status of air conditioners according to claim 8, characterized in that: In step S6, determine the operating state of the central air - conditioning system during the prediction period according to the load rate of the central air - conditioning system obtained in step S5. Specifically, Among the load rates during the prediction period, any one of the following three situations is determined as the startup stage: Situation 1: The load rate shows a continuous increase, and the rising change rate of the load rate at the starting moment and the ending moment is within the first set interval; Situation 2: There is a period during the prediction period when the load rate reaches within the second set interval; Situation 3: The load rate shows a continuous decrease, and the falling change rate of the load rate at the starting moment and the ending moment is within the third set interval; When the load rate during the prediction period remains within the set range and the load rate change rate is within the fourth set interval, it is determined as the running stage; The falling change rate of the load rate at the starting moment and the ending moment is within the fifth set interval, and the load rate drops to 0, which is determined as the stop stage.

Images