Formula Chcd

[Azalee Freas]

Element01 adds the beginning balance field with all of the current year posting amounts for Fixed Assets cost accounts. A cost record in the Fixed Assets Balances table (F1202) has a balance character code (FLCHCD) of 1.

Element 02 uses the beginning balance field for the accumulated depreciation expense accounts to calculate the prior year-ending balance amounts. A primary accumulated depreciation expense account in the Fixed Assets Balances table (F1202) has a balance character code (FLCHCD) of 2.

Calculate results from the formula that is associated with the Salvage Value formula defined in the current life year rule. Use this element in any formula for Upper/Lower Limits, Basis, or depreciation formulas. The default value is zero.

Calculate results from the formula that is associated with the Lower Limit formula defined in the current life year rule. Use the Lower Limit value (Element 08) in any formula for Basis, or Depreciation formulas. The default value is zero.

Calculate results from the formula that is associated with the Upper Limit formula defined in the current life year rule. Use the Upper Limit value (Element 09) in any formula for Basis, or Depreciation formulas. The default value is zero.

Calculate results from the formula that is associated with the Basis formula defined in the current life year rule. Use the Basis value (Element 10) in Depreciation formulas. Default Value is Element 01 (Inception to Date Cost).

This element is derived from the percent amount that you specify in the Annual Multiplier field on the Life Year Rules form (W12851E). Use this element to establish a fixed percentage that can be used in a formula for a specific period of time in an asset's life. If a depreciation formula is not defined, the system automatically uses depreciation formula 95.

This element is derived from the value that is specified in the Multiplier/Constant field in the depreciation formula. To define a constant for Upper/Lower limits, the formula is defined as 12 with the appropriate constant value in the Multiplier/Constant field.

The prior year balance from the Asset Account Balances File table (F1202) for an asset, ledger type, subledger, or subledger type that is related to secondary depreciation accounts. The SDA AAI identifies the secondary accumulated depreciation accounts. The account type of 3 should be used to retrieve the secondary depreciation. (CHCD = 3 in F1202)

This element uses the sum of all of the prior current period posting fields from the Asset Accounts Balances File table (F1202) for an asset, ledger type, subledger, or subledger type. The account type of 3 should be used for retrieval. (CHCD = 3 in F1202)

If the modified start date is not the start of or end of the period, this element calculates the apportionment percentage to us. Mid-Month or Actual Start Dates are examples of number of days that do not match full period results.

Most resources are not case-sensitive (element symbols may be typed in upper or lower case letters). One exception is the Chapman and Hall Chemical Dicitionaries on CD-ROM. The molecular formula indexes in CHCD are case sensitive.

In recent years, the application of deep learning methods in non-invasive load monitoring has shown significant growth. Deep learning methods can understand the deep structures within data, enhancing the performance of models. Additionally, deep learning models are highly flexible in capturing nonlinear relationships within data, which is particularly advantageous for tackling complex problems. Deep learning has become one of the primary applications in the field of load identification. In the majority of non-intrusive load identification algorithms, appliance identification typically occurs when the load category and the number of loads are already known. Guo L, Wang [8] et al. proposed a load identification method based on active deep learning and discrete wavelet transform. Arash [9] et al. proposed a method utilizing a convolutional neural network based on deep learning, employing a layered structure and feature extraction from power consumption curves to achieve appliance type detection and load disaggregation. Weicheng Liu [10] et al. proposed a time-domain power hybrid algorithm and a temporal convolutional autoencoder model to enhance the data processing accuracy of NILM. Eduardo [11] et al. applied the pinball quantile loss function to guide a deep neural network in NILM. Dong Ding [12] et al. proposed a method that utilizes multiple overlapping sliding windows and an improved convolutional neural network internal structure to effectively disaggregate highly mixed loads of multiple appliances. Xiao Zhou [13] et al. employed a deep learning model combining convolutional neural network, long short-term memory network, and random forest algorithms, effectively improving the accuracy of electrical appliance recognition. Leitao Qu [14] et al. proposed a residual convolutional neural network with energy normalization and squeeze-and-excitation blocks, applied in NILM.

But most load identification research currently concentrates on known electrical appliances with identified types and data characteristics. However, there is relatively less research dedicated to identifying unknown loads. Baets [15] proposed a clustering method based on the Siamese neural network for unknown device detection. The advantage of this method is that there is no mandatory limit on the number of equipment types, and there is no need to pre-set the number of clusters. M. Yu [16] et al. proposed a non-invasive load identification model based on the Siamese neural network. The model calculates the similarity of V-I trajectories using the Siamese network and dynamically incorporates new features into the feature library to identify unknown loads. Both methods utilize Siamese neural networks. However, training the Siamese neural network requires pairs of images to determine whether they belong to the same category or not. Moreover, the network may encounter challenges in distinguishing targets with high similarity. Bo Yin [17] et al. utilized the Siamese neural network for unknown device detection based on steady-state single-cycle current. In load identification, V-I trajectory provides more comprehensive information than single cycle current. For example, current waveform, phase information, frequency components, etc. Through these features, it is possible to describe the load behavior more accurately. Additionally, V-I trajectories are very useful for detecting and identifying non-linear loads. The V-I trajectory captures the nonlinear nature of the waveform. A single cycle current may not provide enough information. This method also uses the Siamese neural network and also has the disadvantages of the above two methods. Triplet neural network is typically more suitable for multi-class problems in load identification. It can reduce the labeling data cost, enhance generalization, and adapt to imbalanced data distributions. It effectively addresses the limitations of Siamese neural networks.

The load characteristics can generally be divided into two types: steady-state characteristics [18] and transient characteristics [19] according to the different states of the load. In non-intrusive load identification, the characteristics of the load to be identified are determined by two factors: the electronic components present in the load equipment and its internal circuit structure. The steady-state characteristics and transient characteristics generally mentioned in power system research are more common in fault analysis and diagnosis [18, 20, 21]. Steady-state characteristics are typically more stable and less susceptible to noise and measurement errors, which can result in more reliable outcomes. Measurements of transient characteristics may be influenced by external disturbances. They may also require higher sampling frequencies. Additionally, more complex instrumentation might be needed. These factors have the potential to increase experimental uncertainties. Steady-state data is generally easier to process and analyze since it does not contain momentary fluctuations or noise. This facilitates researchers in extracting valuable information and trends from the data more easily. Different from the characteristics in system research, the non-intrusive load characteristics are more microscopic and correspond to a single load device. In this paper, steady-state features are selected for experiments.

Event detection is required before feature extraction. Understanding the events of electrical equipment under different operating states is crucial. Every time the electrical status changes, an event occurs. Fig 1 is an electrical operation waveform diagram drawn using the root mean square current(RMS) [22]. The red dot is the moment when the event occurs. Under the conditions of determining the turning on or off event of the electrical equipment, the interval range in the stable state is obtained. Anderson [23]et al. provide a framework for evaluating event detection algorithms in non-intrusive load monitoring. The accuracy of event detection is the basis for steady-state feature extraction. After obtaining the stable operation interval of the electrical appliance, we extract various steady-state characteristics of the electrical appliance.

Based on the stable state of operation, each electrical appliance extracts the current and voltage under different complete cycles and then normalizes the extracted data. The calculation formula is described in formula (1) I m ( t ) = I ( t ) - I m i n I m a x - I m i n V m ( t ) = V ( t ) - V m i n V m a x - V m i n (1)In the calculation formula, Im(t) and Vm(t) are the value of the current and voltage after normalization. The unit of Im(t) is A. The unit of Vm(t) is V. I(t) and V(t) are the current and voltage at the present. The unit of I(t) is A. The unit of V(t) is V. Imin and Vmin are the minimum values of current and voltage at the present steady-state cycle. The unit of Imin is A. The unit of Vmin is V. Imax and Vmax are the maximum values of current and voltage at the present steady-state cycle. The unit of Imax is A. The unit of Vmax is V. The value of the minimum and maximum is up to the maximum and minimum values of each electrical device data, not fixed.

3a8082e126